Professional Documents
Culture Documents
P RASAD C HALASANI Carnegie Mellon University chal@cs.cmu.edu S OMESH J HA Carnegie Mellon University sjha@cs.cmu.edu
THIS IS A DRAFT: PLEASE DO NOT DISTRIBUTE c Copyright; Steven E. Shreve, 1996 October 6, 1997
Contents
1 Introduction to Probability Theory 1.1 1.2 1.3 1.4 1.5 The Binomial Asset Pricing Model . . . . . . . . . . . . . . . . . . . . . . . . . . Finite Probability Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lebesgue Measure and the Lebesgue Integral . . . . . . . . . . . . . . . . . . . . General Probability Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Independence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.1 1.5.2 1.5.3 1.5.4 1.5.5 1.5.6 1.5.7 Independence of sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Independence of -algebras . . . . . . . . . . . . . . . . . . . . . . . . . Independence of random variables . . . . . . . . . . . . . . . . . . . . . . Correlation and independence . . . . . . . . . . . . . . . . . . . . . . . . Independence and conditional expectation. . . . . . . . . . . . . . . . . . Law of Large Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . Central Limit Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 11 16 22 30 40 40 41 42 44 45 46 47 49 49 50 52 52 53 54 55 57 58
2 Conditional Expectation 2.1 2.2 2.3 A Binomial Model for Stock Price Dynamics . . . . . . . . . . . . . . . . . . . . Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conditional Expectation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.1 2.3.2 2.3.3 2.3.4 2.3.5 2.4 An example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Denition of Conditional Expectation . . . . . . . . . . . . . . . . . . . . Further discussion of Partial Averaging . . . . . . . . . . . . . . . . . . . Properties of Conditional Expectation . . . . . . . . . . . . . . . . . . . . Examples from the Binomial Model . . . . . . . . . . . . . . . . . . . . .
Martingales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 3 Arbitrage Pricing 3.1 3.2 3.3 Binomial Pricing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . General one-step APT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Risk-Neutral Probability Measure . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1 3.3.2 3.4 3.5 Portfolio Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Self-nancing Value of a Portfolio Process . . . . . . . . . . . . . . . . 59 59 60 61 62 62 63 64 67 67 69 70 70 73 74 77 77 79 81 85 85 86 88 89 91 91 92 94 97 97
4 The Markov Property 4.1 4.2 4.3 4.4 4.5 Binomial Model Pricing and Hedging . . . . . . . . . . . . . . . . . . . . . . . . Computational Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Markov Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Different ways to write the Markov property . . . . . . . . . . . . . . . . Showing that a process is Markov . . . . . . . . . . . . . . . . . . . . . . . . . . Application to Exotic Options . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5 Stopping Times and American Options 5.1 5.2 5.3 American Pricing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Value of Portfolio Hedging an American Option . . . . . . . . . . . . . . . . . . . Information up to a Stopping Time . . . . . . . . . . . . . . . . . . . . . . . . . .
6 Properties of American Derivative Securities 6.1 6.2 6.3 6.4 The properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Proofs of the Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Compound European Derivative Securities . . . . . . . . . . . . . . . . . . . . . . Optimal Exercise of American Derivative Security . . . . . . . . . . . . . . . . . .
7 Jensens Inequality 7.1 7.2 7.3 Jensens Inequality for Conditional Expectations . . . . . . . . . . . . . . . . . . . Optimal Exercise of an American Call . . . . . . . . . . . . . . . . . . . . . . . . Stopped Martingales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3 8.2 8.3 8.4 8.5 8.6 8.7 8.8 8.9 is almost surely nite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The moment generating function for Expectation of . . . . . . . . . . . . . . . . . . . . . . . . 97 99
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
The Strong Markov Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 General First Passage Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 Example: Perpetual American Put . . . . . . . . . . . . . . . . . . . . . . . . . . 102 Difference Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 Distribution of First Passage Times . . . . . . . . . . . . . . . . . . . . . . . . . . 107
8.10 The Reection Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 9 Pricing in terms of Market Probabilities: The Radon-Nikodym Theorem. 9.1 9.2 9.3 9.4 9.5 111
Radon-Nikodym Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 Radon-Nikodym Martingales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 The State Price Density Process . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 Stochastic Volatility Binomial Model . . . . . . . . . . . . . . . . . . . . . . . . . 116 Another Applicaton of the Radon-Nikodym Theorem . . . . . . . . . . . . . . . . 118 119
11.1 Law of a Random Variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 11.2 Density of a Random Variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 11.3 Expectation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 11.4 Two random variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 11.5 Marginal Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126 11.6 Conditional Expectation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126 11.7 Conditional Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 11.8 Multivariate Normal Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 11.9 Bivariate normal distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 11.10MGF of jointly normal random variables . . . . . . . . . . . . . . . . . . . . . . . 130 12 Semi-Continuous Models 131
4 12.2 The Stock Price Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132 12.3 Remainder of the Market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 12.4 Risk-Neutral Measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 12.5 Risk-Neutral Pricing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 12.6 Arbitrage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 12.7 Stalking the Risk-Neutral Measure . . . . . . . . . . . . . . . . . . . . . . . . . . 135 12.8 Pricing a European Call . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 13 Brownian Motion 139
13.1 Symmetric Random Walk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 13.2 The Law of Large Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 13.3 Central Limit Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 13.4 Brownian Motion as a Limit of Random Walks . . . . . . . . . . . . . . . . . . . 141 13.5 Brownian Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142 13.6 Covariance of Brownian Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . 143 13.7 Finite-Dimensional Distributions of Brownian Motion . . . . . . . . . . . . . . . . 144 13.8 Filtration generated by a Brownian Motion . . . . . . . . . . . . . . . . . . . . . . 144 13.9 Martingale Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 13.10The Limit of a Binomial Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 13.11Starting at Points Other Than 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 13.12Markov Property for Brownian Motion . . . . . . . . . . . . . . . . . . . . . . . . 147 13.13Transition Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 13.14First Passage Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 14 The It Integral o 153
14.1 Brownian Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 14.2 First Variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 14.3 Quadratic Variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155 14.4 Quadratic Variation as Absolute Volatility . . . . . . . . . . . . . . . . . . . . . . 157 14.5 Construction of the It Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 o 14.6 It integral of an elementary integrand . . . . . . . . . . . . . . . . . . . . . . . . 158 o 14.7 Properties of the It integral of an elementary process . . . . . . . . . . . . . . . . 159 o 14.8 It integral of a general integrand . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 o
5 14.9 Properties of the (general) It integral . . . . . . . . . . . . . . . . . . . . . . . . 163 o 14.10Quadratic variation of an It integral . . . . . . . . . . . . . . . . . . . . . . . . . 165 o 15 It s Formula o 167
15.1 It s formula for one Brownian motion . . . . . . . . . . . . . . . . . . . . . . . . 167 o 15.2 Derivation of It s formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168 o 15.3 Geometric Brownian motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 15.4 Quadratic variation of geometric Brownian motion . . . . . . . . . . . . . . . . . 170 15.5 Volatility of Geometric Brownian motion . . . . . . . . . . . . . . . . . . . . . . 170 15.6 First derivation of the Black-Scholes formula . . . . . . . . . . . . . . . . . . . . 170 15.7 Mean and variance of the Cox-Ingersoll-Ross process . . . . . . . . . . . . . . . . 172 15.8 Multidimensional Brownian Motion . . . . . . . . . . . . . . . . . . . . . . . . . 173 15.9 Cross-variations of Brownian motions . . . . . . . . . . . . . . . . . . . . . . . . 174 15.10Multi-dimensional It formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 o 16 Markov processes and the Kolmogorov equations 177
16.1 Stochastic Differential Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 16.2 Markov Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178 16.3 Transition density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179 16.4 The Kolmogorov Backward Equation . . . . . . . . . . . . . . . . . . . . . . . . 180 16.5 Connection between stochastic calculus and KBE . . . . . . . . . . . . . . . . . . 181 16.6 Black-Scholes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 16.7 Black-Scholes with price-dependent volatility . . . . . . . . . . . . . . . . . . . . 186 17 Girsanovs theorem and the risk-neutral measure 17.1 Conditional expectations under
f IP
189
. . . . . . . . . . . . . . . . . . . . . . . . . . 191
18.1 Martingale Representation Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 197 18.2 A hedging application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 18.3 18.4
. . . . . . . . . . . . . . . 199 . . . . . . . . . . . . . . . 200
6 19 A two-dimensional market model 19.1 Hedging when ;1 < 19.2 Hedging when 203 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
=1
<1 .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 209
20.1 Reection principle for Brownian motion . . . . . . . . . . . . . . . . . . . . . . 209 20.2 Up and out European call. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212 20.3 A practical issue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218 21 Asian Options 219
21.1 Feynman-Kac Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220 21.2 Constructing the hedge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220 21.3 Partial average payoff Asian option . . . . . . . . . . . . . . . . . . . . . . . . . . 221 22 Summary of Arbitrage Pricing Theory 223
22.1 Binomial model, Hedging Portfolio . . . . . . . . . . . . . . . . . . . . . . . . . 223 22.2 Setting up the continuous model . . . . . . . . . . . . . . . . . . . . . . . . . . . 225 22.3 Risk-neutral pricing and hedging . . . . . . . . . . . . . . . . . . . . . . . . . . . 227 22.4 Implementation of risk-neutral pricing and hedging . . . . . . . . . . . . . . . . . 229 23 Recognizing a Brownian Motion 233
23.1 Identifying volatility and correlation . . . . . . . . . . . . . . . . . . . . . . . . . 235 23.2 Reversing the process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236 24 An outside barrier option 239
24.1 Computing the option value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242 24.2 The PDE for the outside barrier option . . . . . . . . . . . . . . . . . . . . . . . . 243 24.3 The hedge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245 25 American Options 247
25.1 Preview of perpetual American put . . . . . . . . . . . . . . . . . . . . . . . . . . 247 25.2 First passage times for Brownian motion: rst method . . . . . . . . . . . . . . . . 247 25.3 Drift adjustment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249 25.4 Drift-adjusted Laplace transform . . . . . . . . . . . . . . . . . . . . . . . . . . . 250 25.5 First passage times: Second method . . . . . . . . . . . . . . . . . . . . . . . . . 251
7 25.6 Perpetual American put . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252 25.7 Value of the perpetual American put . . . . . . . . . . . . . . . . . . . . . . . . . 256 25.8 Hedging the put . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257 25.9 Perpetual American contingent claim . . . . . . . . . . . . . . . . . . . . . . . . . 259 25.10Perpetual American call . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259 25.11Put with expiration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260 25.12American contingent claim with expiration . . . . . . . . . . . . . . . . . . . . . 261 26 Options on dividend-paying stocks 263
26.1 American option with convex payoff function . . . . . . . . . . . . . . . . . . . . 263 26.2 Dividend paying stock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264 26.3 Hedging at time t1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266 267
27.1 Forward contracts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269 27.2 Hedging a forward contract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269 27.3 Future contracts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270 27.4 Cash ow from a future contract . . . . . . . . . . . . . . . . . . . . . . . . . . . 272 27.5 Forward-future spread . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272 27.6 Backwardation and contango . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273 28 Term-structure models 275
28.1 Computing arbitrage-free bond prices: rst method . . . . . . . . . . . . . . . . . 276 28.2 Some interest-rate dependent assets . . . . . . . . . . . . . . . . . . . . . . . . . 276 28.3 Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277 28.4 Forward rate agreement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277 28.5 Recovering the interest r(t) from the forward rate . . . . . . . . . . . . . . . . . . 278 28.6 Computing arbitrage-free bond prices: Heath-Jarrow-Morton method . . . . . . . . 279 28.7 Checking for absence of arbitrage . . . . . . . . . . . . . . . . . . . . . . . . . . 280 28.8 Implementation of the Heath-Jarrow-Morton model . . . . . . . . . . . . . . . . . 281 29 Gaussian processes 285
29.1 An example: Brownian Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286 30 Hull and White model 293
8 30.1 Fiddling with the formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295 30.2 Dynamics of the bond price . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296 30.3 Calibration of the Hull & White model . . . . . . . . . . . . . . . . . . . . . . . . 297 30.4 Option on a bond . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299 31 Cox-Ingersoll-Ross model 303
31.1 Equilibrium distribution of r(t) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306 31.2 Kolmogorov forward equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306 31.3 Cox-Ingersoll-Ross equilibrium density . . . . . . . . . . . . . . . . . . . . . . . 309 31.4 Bond prices in the CIR model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310 31.5 Option on a bond . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313 31.6 Deterministic time change of CIR model . . . . . . . . . . . . . . . . . . . . . . . 313 31.7 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315 31.8 Tracking down '0(0) in the time change of the CIR model . . . . . . . . . . . . . 316 32 A two-factor model (Dufe & Kan) 319
32.1 Non-negativity of Y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320 32.2 Zero-coupon bond prices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321 32.3 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323 33 Change of num raire e 325
33.1 Bond price as num raire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327 e 33.2 Stock price as num raire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328 e 33.3 Merton option pricing formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329 34 Brace-Gatarek-Musiela model 335
34.1 Review of HJM under risk-neutral IP . . . . . . . . . . . . . . . . . . . . . . . . . 335 34.2 Brace-Gatarek-Musiela model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336 34.3 LIBOR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337 34.4 Forward LIBOR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338 34.5 The dynamics of L(t
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
34.6 Implementation of BGM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340 34.7 Bond prices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342 34.8 Forward LIBOR under more forward measure . . . . . . . . . . . . . . . . . . . . 343
9 34.9 Pricing an interest rate caplet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343 34.10Pricing an interest rate cap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345 34.11Calibration of BGM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345 34.12Long rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346 34.13Pricing a swap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346 35 Notes and References 349
35.1 Probability theory and martingales. . . . . . . . . . . . . . . . . . . . . . . . . . . 349 35.2 Binomial asset pricing model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349 35.3 Brownian motion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350 35.4 Stochastic integrals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350 35.5 Stochastic calculus and nancial markets. . . . . . . . . . . . . . . . . . . . . . . 350 35.6 Markov processes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351 35.7 Girsanovs theorem, the martingale representation theorem, and risk-neutral measures.351 35.8 Exotic options. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352 35.9 American options. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352 35.10Forward and futures contracts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353 35.11Term structure models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353 35.12Change of num raire. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353 e 35.13Foreign exchange models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353 35.14REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
10
Chapter 1
0<d<u
(1.1)
such that at the next period, the stock price will be either dS 0 or uS0. Typically, we take d and u to satisfy 0 < d < 1 < u, so change of the stock price from S0 to dS0 represents a downward movement, and change of the stock price from S0 to uS0 represents an upward movement. It is 1 common to also have d = u , and this will be the case in many of our examples. However, strictly speaking, for what we are about to do we need to assume only (1.1) and (1.2) below. Of course, stock price movements are much more complicated than indicated by the binomial asset pricing model. We consider this simple model for three reasons. First of all, within this model the concept of arbitrage pricing and its relation to risk-neutral pricing is clearly illuminated. Secondly, the model is used in practice because with a sufcient number of steps, it provides a good, computationally tractable approximation to continuous-time models. Thirdly, within the binomial model we can develop the theory of conditional expectations and martingales which lies at the heart of continuous-time models. With this third motivation in mind, we develop notation for the binomial model which is a bit different from that normally found in practice. Let us imagine that we are tossing a coin, and when we get a Head, the stock price moves up, but when we get a Tail, the price moves down. We denote the price at time 1 by S1 (H ) = uS0 if the toss results in head (H), and by S 1 (T ) = dS0 if it 11
12
S2 (HH) = 16
S2 (TT) = 1
Figure 1.1: Binomial tree of stock prices with S 0 results in tail (T). After the second toss, the price will be one of:
= 4, u = 1=d = 2.
S2 (HH ) = uS1(H ) = u2 S0 S2(HT ) = dS1(H ) = duS0 S2 (TH ) = uS1(T ) = udS0 S2(TT ) = dS1(T ) = d2S0:
After three tosses, there are eight possible coin sequences, although not all of them result in different stock prices at time 3. For the moment, let us assume that the third toss is the last one and denote by
The stock price Sk at time k depends on the coin tosses. To emphasize this, we often write S k (! ). Actually, this notation does not quite tell the whole story, for while S 3 depends on all of ! , S2 depends on only the rst two components of ! , S 1 depends on only the rst component of ! , and S0 does not depend on ! at all. Sometimes we will use notation such S 2(!1 !2) just to record more explicitly how S 2 depends on ! = (!1 !2 !3 ).
1 Example 1.1 Set S0 = 4, u = 2 and d = 2 . We have then the binomial tree of possible stock prices shown in Fig. 1.1. Each sample point ! = (! 1 !2 !3) represents a path through the tree. Thus, we can think of the sample space as either the set of all possible outcomes from three coin tosses or as the set of all possible paths through the tree.
To complete our binomial asset pricing model, we introduce a money market with interest rate r; $1 invested in the money market becomes $(1 + r) in the next period. We take r to be the interest
13
rate for both borrowing and lending. (This is not as ridiculous as it rst seems, because in a many applications of the model, an agent is either borrowing or lending (not both) and knows in advance which she will be doing; in such an application, she should take r to be the rate of interest for her activity.) We assume that
d < 1 + r < u:
(1.2)
The model would not make sense if we did not have this condition. For example, if 1 + r u, then the rate of return on the money market is always at least as great as and sometimes greater than the return on the stock, and no one would invest in the stock. The inequality d 1 + r cannot happen unless either r is negative (which never happens, except maybe once upon a time in Switzerland) or d 1. In the latter case, the stock does not really go down if we get a tail; it just goes up less than if we had gotten a head. One should borrow money at interest rate r and invest in the stock, since even in the worst case, the stock price rises at least as fast as the debt used to buy it.
K > 0 and expiration time 1. This option confers the right to buy the stock at time 1 for K dollars, and so is worth S 1 ; K at time 1 if S1 ; K is positive and is otherwise worth zero. We denote by V1(!) = (S1(! ) ; K )+ = maxfS1(!) ; K 0g the value (payoff) of this option at expiration. Of course, V 1(! ) actually depends only on ! 1 , and we can and do sometimes write V1 (!1) rather than V1(! ). Our rst task is to compute the arbitrage
price of this option at time zero. Suppose at time zero you sell the call for V0 dollars, where V0 is still to be determined. You now have an obligation to pay off (uS 0 ; K )+ if !1 = H and to pay off (dS0 ; K )+ if !1 = T . At the time you sell the option, you dont yet know which value ! 1 will take. You hedge your short position in the option by buying 0 shares of stock, where 0 is still to be determined. You can use the proceeds V0 of the sale of the option for this purpose, and then borrow if necessary at interest rate r to complete the purchase. If V0 is more than necessary to buy the 0 shares of stock, you invest the residual money at interest rate r. In either case, you will have V 0 ; 0S0 dollars invested in the money market, where this quantity might be negative. You will also own 0 shares of stock. If the stock goes up, the value of your portfolio (excluding the short position in the option) is
With the stock as the underlying asset, let us consider a European call option with strike price
0 S1(H ) + (1 + r)(V0 ; 0 S0 )
and you need to have V1(H ). Thus, you want to choose V 0 and
0 so that
(1.3)
V1(H ) =
0 S1 (T ) + (1 + r)(V0 ; 0 S0)
and you need to have V1(T ). Thus, you want to choose V 0 and
0 to also have
(1.4)
V 1 (T ) =
14 These are two equations in two unknowns, and we solve them below Subtracting (1.4) from (1.3), we obtain
V1 (H ) ; V1(T ) =
so that
0 (S1(H ) ; S1 (T ))
(1.5)
0=
(1.6)
This is a discrete-time version of the famous delta-hedging formula for derivative securities, according to which the number of shares of an underlying asset a hedge should hold is the derivative (in the sense of calculus) of the value of the derivative security with respect to the price of the underlying asset. This formula is so pervasive the when a practitioner says delta, she means the derivative (in the sense of calculus) just described. Note, however, that my denition of 0 is the number of shares of stock one holds at time zero, and (1.6) is a consequence of this denition, not the denition of 0 itself. Depending on how uncertainty enters the model, there can be cases in which the number of shares of stock a hedge should hold is not the (calculus) derivative of the derivative security with respect to the price of the underlying asset. To complete the solution of (1.3) and (1.4), we substitute (1.6) into either (1.3) or (1.4) and solve for V0 . After some simplication, this leads to the formula
1 r V0 = 1 + r 1 + ; ; d V1(H ) + u ; (1 + r) V1 (T ) : u d u;d This is the arbitrage price for the European call option with payoff V 1 at time 1.
formula, we dene
r p = 1 + ;; d ~ u d
so that (1.7) becomes
q = u ; (1 + r) = 1 ; p ~ ~ u;d
(1.8)
(1.9)
Because we have taken d < u, both p and q are dened,i.e., the denominator in (1.8) is not zero. ~ ~ Because of (1.2), both p and q are in the interval (0 1), and because they sum to 1, we can regard ~ ~ them as probabilities of H and T , respectively. They are the risk-neutral probabilites. They appeared when we solved the two equations (1.3) and (1.4), and have nothing to do with the actual probabilities of getting H or T on the coin tosses. In fact, at this point, they are nothing more than a convenient tool for writing (1.7) as (1.9). We now consider a European call which pays off K dollars at time 2. At expiration, the payoff of this option is V 2 = (S2 ; K )+ , where V2 and S2 depend on !1 and !2 , the rst and second coin tosses. We want to determine the arbitrage price for this option at time zero. Suppose an agent sells the option at time zero for V0 dollars, where V0 is still to be determined. She then buys 0 shares
15
of stock, investing V 0 ; 0S0 dollars in the money market to nance this. At time 1, the agent has a portfolio (excluding the short position in the option) valued at
X1 =
0 S1 + (1 + r)(V0 ; 0 S0):
(1.10)
Although we do not indicate it in the notation, S 1 and therefore X1 depend on !1 , the outcome of the rst coin toss. Thus, there are really two equations implicit in (1.10):
X1(H ) = X1(T ) =
0S1 (T ) + (1 + r)(V0 ; 0 S0 ):
After the rst coin toss, the agent has X 1 dollars and can readjust her hedge. Suppose she decides to now hold 1 shares of stock, where 1 is allowed to depend on ! 1 because the agent knows what value !1 has taken. She invests the remainder of her wealth, X1 ; 1S1 in the money market. In the next period, her wealth will be given by the right-hand side of the following equation, and she wants it to be V 2. Therefore, she wants to have
V2 =
1 S2 + (1 + r)(X1 ; 1 S1):
(1.11)
Although we do not indicate it in the notation, S 2 and V2 depend on !1 and !2 , the outcomes of the rst two coin tosses. Considering all four possible outcomes, we can write (1.11) as four equations:
= = = =
1(H )S2(HH ) + (1 + r)(X1(H ) ; 1 (H )S1(H )) 1(H )S2(HT ) + (1 + r)(X1(H ) ; 1(H )S1(H )) 1(T )S2(TH ) + (1 + r)(X1(T ) ; 1(T )S1(T )) 1(T )S2(TT ) + (1 + r)(X1(T ) ; 1 (T )S1(T )):
To solve these equations, and thereby determine the arbitrage price V 0 at time zero of the option and the hedging portfolio 0 , 1(H ) and 1 (T ), we begin with the last two
We now have six equations, the two represented by (1.10) and the four represented by (1.11), in the six unknowns V 0 , 0 , 1(H ), 1 (T ), X1 (H ), and X1 (T ).
V2(TH ) = V2(TT ) =
1 (T )S2(TH ) + (1 + r)(X1(T ) ; 1(T )S1(T )) 1 (T )S2(TT ) + (1 + r)(X1(T ) ; 1 (T )S1(T )): 1(T ), we obtain the delta-hedging for(1.12)
Subtracting one of these from the other and solving for mula
1(T ) =
(1.13)
16 Equation (1.13), gives the value the hedging portfolio should have at time 1 if the stock goes down between times 0 and 1. We dene this quantity to be the arbitrage value of the option at time 1 if !1 = T , and we denote it by V1(T ). We have just shown that
(1.14)
The hedger should choose her portfolio so that her wealth X 1 (T ) if !1 = T agrees with V1(T ) dened by (1.14). This formula is analgous to formula (1.9), but postponed by one step. The rst two equations implicit in (1.11) lead in a similar way to the formulas
1 (H ) =
(1.15)
and X1(H ) = V1(H ), where V1(H ) is the value of the option at time 1 if ! 1
= H , dened by
(1.16)
This is again analgous to formula (1.9), postponed by one step. Finally, we plug the values X 1(H ) = V1(H ) and X1(T ) = V1(T ) into the two equations implicit in (1.10). The solution of these equations for 0 and V0 is the same as the solution of (1.3) and (1.4), and results again in (1.6) and (1.9). The pattern emerging here persists, regardless of the number of periods. If Vk denotes the value at time k of a derivative security, and this depends on the rst k coin tosses ! 1 : : : !k , then at time k ; 1, after the rst k ; 1 tosses !1 : : : !k;1 are known, the portfolio to hedge a short position should hold k;1 (!1 : : : !k;1) shares of stock, where
H ) ; Vk (!1 : : : !k 1 T ) (1.17) H ) ; Sk (!1 : : : !k 1 T ) and the value at time k ; 1 of the derivative security, when the rst k ; 1 coin tosses result in the outcomes !1 : : : !k 1 , is given by
k 1 (!1
;
V : : : !k 1 ) = Sk (!1 :: :: :: !k !k k (!1
;
1 1
H ) + q Vk (!1 : : : !k ~
T )]
(1.18)
(2.1)
of all possible outcomes of three coin tosses. Let F be the set of all subsets of . Some sets in F are , fHHH HHT HTH HTT g, fTTT g, and itself. How many sets are there in F ?
17
IP
is a function mapping
k=1
Ak =
! X
1
k=1
IP (Ak ):
Probability measures have the following interpretation. Let A be a subset of F . Imagine that is the set of all possible outcomes of some random experiment. There is a certain probability, between 0 and 1, that when that experiment is performed, the outcome will lie in the set A. We think of IP (A) as this probability. Example 1.2 Suppose a coin has probability 1 for H and 2 for T . For the individual elements of 3 3 in (2.1), dene
1 3 1 3 1 3 1 3
3 2 2 3 2 1 3 2 2 3
1 2 2 3 3 1 2 2 3 3 1 2 2 3 3 2 3: 3
IP (A) =
For example,
X
! A
2
IP f!g:
2
(2.2)
2 + 1 3 3
2 3
=1 3
which is another way of saying that the probability of H on the rst toss is 1 . 3 As in the above example, it is generally the case that we specify a probability measure on only some of the subsets of and then use property (ii) of Denition 1.1 to determine IP (A) for the remaining sets A 2 F . In the above example, we specied the probability measure only for the sets containing a single element, and then used Denition 1.1(ii) in the form (2.2) (see Problem 1.4(ii)) to determine IP for all the other sets in F . Denition 1.2 Let be a nonempty set. A -algebra is a collection following three properties: (i)
G of subsets of
with the
2 G,
2 G,
k=1 Ak is also in G .
1
F0 = F1 = F2 =
( (
in Example 1.2:
) fHHH HHT HTH HTT g fTHH THT TTH TTT g fHHH HHT g fHTH HTT g fTHH THT g fTTH TTT g )
and all sets which can be built by taking unions of these
AH = fHHH HHT HTH HTT g = fH on the rst tossg AT = fTHH THT TTH TTT g = fT on the rst tossg
so that and let us dene
F1 = f
AH AT g
AHH = fHHH HHT g = fHH on the rst two tossesg AHT = fHTH HTT g = fHT on the rst two tossesg ATH = fTHH THT g = fTH on the rst two tossesg ATT = fTTH TTT g = fTT on the rst two tossesg
so that
F2 = f
AHH AHT ATH ATT AH AT AHH ATH AHH ATT AHT ATH AHT ATT Ac Ac Ac Ac g: HH HT TH TT
We interpret -algebras as a record of information. Suppose the coin is tossed three times, and you are not told the outcome, but you are told, for every set in F 1 whether or not the outcome is in that set. For example, you would be told that the outcome is not in and is in . Moreover, you might be told that the outcome is not in A H but is in A T . In effect, you have been told that the rst toss was a T , and nothing more. The -algebra F1 is said to contain the information of the rst toss, which is usually called the information up to time 1. Similarly, F 2 contains the information of
19
the rst two tosses, which is the information up to time 2. The -algebra F 3 = F contains full information about the outcome of all three tosses. The so-called trivial -algebra F 0 contains no information. Knowing whether the outcome ! of the three tosses is in (it is not) and whether it is in (it is) tells you nothing about ! Denition 1.3 Let be a nonempty nite set. A ltration is a sequence of -algebras F 0 F1 F2 : : : such that each -algebra in the sequence contains all the sets contained by the previous -algebra. Denition 1.4 Let be a nonempty nite set and let random variable is a function mapping into IR.
Fn
F be the
. A
Example 1.3 Let be given by (2.1) and consider the binomial asset pricing Example 1.1, where S0 = 4, u = 2 and d = 1 . Then S0, S1, S2 and S3 are all random variables. For example, 2 S2 (HHT ) = u2 S0 = 16. The random variable S0 is really not random, since S0(!) = 4 for all ! 2 . Nonetheless, it is a function mapping into IR, and thus technically a random variable, albeit a degenerate one. A random variable maps into IR, and we can look at the preimage under the random variable of sets in IR. Consider, for example, the random variable S2 of Example 1.1. We have
20 Denition 1.5 Let be a nonemtpy nite set and let F be the -algebra of all subsets of . Let X be a random variable on ( F ). The -algebra (X ) generated by X is dened to be the collection of all sets of the form f! 2 X (! ) 2 Ag, where A is a subset of IR. Let G be a sub- -algebra of F . We say that X is G -measurable if every set in (X ) is also in G . Note: We normally write simply fX
2 Ag rather than f! 2 X (!) 2 Ag. Denition 1.6 Let be a nonempty, nite set, let F be the -algebra of all subsets of , let IP be a probabilty measure on ( F ), and let X be a random variable on . Given any set A IR, we
dene the induced measure of A to be
LX (A) = IP fX 2 Ag:
In other words, the induced measure of a set A tells us the probability that X takes a value in A. In the case of S2 above with the probability measure of Example 1.2, some sets in IR and their induced measures are:
(2.3)
large they are, or tell what the cumulative distribution function is. (Later we will consider random variables X which have densities, in which case the induced measure of a set A IR is the integral of the density over the set A.)
LX . If X is discrete, as in the case of S2 above, we can either tell where the masses are and how
Important Note. In order to work through the concept of a risk-neutral measure, we set up the denitions to make a clear distinction between random variables and their distributions. A random variable is a mapping from to IR, nothing more. It has an existence quite apart from discussion of probabilities. For example, in the discussion above, S 2 (TTH ) = S2(TTT ) = 1, regardless of whether the probability for H is 1 or 1 . 3 2
21
The distribution of a random variable is a measure L X on IR, i.e., a way of assigning probabilities to sets in IR. It depends on the random variable X and the probability measure IP we use in . If we set the probability of H to be 1 , then LS2 assigns mass 1 to the number 16. If we set the probability 3 9 of H to be 1 , then LS2 assigns mass 1 to the number 16. The distribution of S 2 has changed, but 2 4 the random variable has not. It is still dened by
IEX =
X
2
X (!)IP f!g:
(2.4) is a into
Notice that the expected value in (2.4) is dened to be a sum over the sample space . Since nite set, X can take only nitely many values, which we label x 1 : : : xn . We can partition the subsets fX 1 = x1 g : : : fXn = xn g, and then rewrite (2.4) as
IEX =
= = = =
X
2
! n X
X (!)IP f!g
X
2f g
X (!)IP f! g
22 Thus, although the expected value is dened as a sum over the sample space , we can also write it as a sum over IR.
To make the above set of equations absolutely clear, we consider S 2 with the distribution given by (2.3). The denition of IES2 is
IES2 = S2 (HHH )IP fHHH g + S2(HHT )IP fHHT g +S2 (HTH )IP fHTH g + S2(HTT )IP fHTT g +S2 (THH )IP fTHH g + S2(THT )IP fTHT g +S2 (TTH )IP fTTH g + S2 (TTT )IP fTTT g = 16 IP (AHH ) + 4 IP (AHT ATH ) + 1 IP (ATT ) = 16 IP fS2 = 16g + 4 IP fS2 = 4g + 1 IP fS2 = 1g
= 16 LS2 f16g + 4 LS2 f4g + 1 LS2 f1g = 16 1 + 4 4 + 4 4 9 9 9 48 : = 9
Denition 1.8 Let be a nonempty, nite set, let F be the -algebra of all subsets of , let IP be a probabilty measure on ( F ), and let X be a random variable on . The variance of X is dened to be the expected value of (X ; IEX )2, i.e., Var(X ) =
X
2
(X (! ) ; IEX )2IP f! g:
(2.5)
x1 : : : xn, then
One again, we can rewrite (2.5) as a sum over IR rather than over . Indeed, if X takes the values Var(X ) =
n X k=1
n X k=1
23
Denition 1.9 The Borel -algebra, denoted B(IR), is the smallest -algebra containing all open intervals in IR. The sets in B(IR) are called Borel sets. Every set which can be written down and just about every set imaginable is in B(IR). The following discussion of this fact uses the -algebra properties developed in Problem 1.3.
By denition, every open interval (a b) is in B(IR), where a and b are real numbers. Since B(IR) is a -algebra, every union of open intervals is also in B(IR). For example, for every real number a, the open half-line
(a 1) =
is a Borel set, as is
n=1
1
(a a + n) (a ; n a):
(;1 a) =
For real numbers a and b, the union
n=1
(;1 a) (b 1)
is Borel. Since B(IR) is a -algebra, every complement of a Borel set is Borel, so B(IR) contains
a b] = (;1 a) (b 1) :
This shows that every closed interval is Borel. In addition, the closed half-lines
a 1) =
and
n=1
1
a a + n] a ; n a]
(;1 a] =
n=1
are Borel. Half-open and half-closed intervals are also Borel, since they can be written as intersections of open half-lines and closed half-lines. For example,
(a b] = (;1 b] \ (a 1):
Every set which contains only one real number is Borel. Indeed, if a is a real number, then
fag =
\
1
n=1
1 1 a; n a+ n : = fa 1 a2 : : : an g,
n
This means that every set containing nitely many real numbers is Borel; if A then
A=
k=1
fak g:
A=
a2 : : : g, then
k=1
fak g:
This means that the set of rational numbers is Borel, as is its complement, the set of irrational numbers. There are, however, sets which are not Borel. We have just seen that any non-Borel set must have uncountably many points. Example 1.4 (The Cantor set.) This example gives a hint of how complicated a Borel set can be. We use it later when we discuss the sample space for an innite sequence of coin tosses. Consider the unit interval
0 1], and remove the middle half, i.e., remove the open interval 3 A1 = 1 4 : 4
C1 = 0 1 4
1 3 A2 = 16 16
31 4 13 15 : 16 16
has two pieces. From each of these pieces, remove the middle half, i.e., remove the open set
15 16 1 : has four pieces. Continue this process, so at stage k, the set C k has 2k pieces, and each piece has
1 C2 = 0 16
3 1 16 4
3 13 4 16
C=
\
1
k=1
Ck
is dened to be the set of points not removed at any stage of this nonterminating process. Note that the length of A 1 , the rst set removed, is 1 . The length of A2 , the second set removed, 2 1 1 is 1 + 1 = 4 . The length of the next set removed is 4 32 = 1 , and in general, the length of the 8 8 8 ;k k-th set removed is 2 . Thus, the total length removed is
k=1
X1 2k = 1
1
and so the Cantor set, the set of points not removed, has zero length. Despite the fact that the Cantor set has no length, there are lots of points in this set. In particular, none of the endpoints of the pieces of the sets C 1 C2 : : : is ever removed. Thus, the points are all in C . This is a countably innite set of points. We shall see eventually that the Cantor set has uncountably many points.
3 1 3 13 1 0 1 4 1 16 16 16 15 64 : : : 4 16
25
B(IR)) is a
( ) = 0,
(ii) If A1
k=1
Ak =
! X
1
k=1
(Ak ):
Lebesgue measure is dened to be the measure on (IR B(IR)) which assigns the measure of each interval to be its length. Following Williamss book, we denote Lebesgue measure by 0 . A measure has all the properties of a probability measure given in Problem 1.4, except that the total measure of the space is not necessarily 1 (in fact, 0 (IR) = 1), one no longer has the equation
(Ac ) = 1 ; (A)
in Problem 1.4(iii), and property (v) in Problem 1.4 needs to be modied to say: (v) If A1
\
1
k=1
To see that the additional requirment
Ak = nlim (An ):
!1
and
A1 = 1 1) A2 = 2 1) A3 = 3 1) : : ::
Then \1 Ak k=1
We specify that the Lebesgue measure of each interval is its length, and that determines the Lebesgue measure of all other Borel sets. For example, the Lebesgue measure of the Cantor set in Example 1.4 must be zero, because of the length computation given at the end of that example. The Lebesgue measure of a set containing only one point must be zero. In fact, since
fag
for every positive integer n, we must have
1 1 a; n a+ n
0
0
Letting n ! 1, we obtain
0 fag
1 1 2 a ; n a + n = n:
0 fag = 0:
26 The Lebesgue measure of a set containing countably many points must also be zero. Indeed, if A = fa1 a2 : : : g, then
0 (A) =
X
1
k=1
0 fak g =
X
1
k=1
0 = 0:
The Lebesgue measure of a set containing uncountably many points can be either zero, positive and nite, or innite. We may not compute the Lebesgue measure of an uncountable set by adding up the Lebesgue measure of its individual members, because there is no way to add up uncountably many numbers. The integral was invented to get around this problem. In order to think about Lebesgue integrals, we must rst consider the functions to be integrated. Denition 1.11 Let
fx 2 IR f (x) 2 Ag is in B(IR) whenever A 2 B(IR). In the language of Section 2, we want the -algebra generated by f to be contained in B(IR).
Denition 3.4 is purely technical and has nothing to do with keeping track of information. It is difcult to conceive of a function which is not Borel-measurable, and we shall pretend such functions dont exist. Hencefore, function mapping IR to IR will mean Borel-measurable function mapping IR to IR and subset of IR will mean Borel subset of IR. Denition 1.12 An indicator function g from and 1. We call
We say that
IR
g d 0 = 0 (A):
n X
A simple function h from IR to IR is a linear combination of indicators, i.e., a function of the form
h(x) =
where each gk is of the form
k=1
ck gk (x)
if x 2 Ak if x 2 Ak =
gk (x) = hd 0 = ck
1 0
n X Z
k=1
IR
gk d 0 =
n X
k=1
ck 0 (Ak ):
We
IR
f d 0 = sup
IR
hd
27
Finally, let f be a function dened on IR, possibly taking the value 1 at some points and the value ;1 at other points. We dene the positive and negative parts of f to be
f + (x) = maxff (x) 0g f (x) = maxf;f (x) 0g respectively, and we dene the Lebesgue integral of f to be
;
Let f be a function dened on IR, possibly taking the value 1 at some points and the value ;1 at other points. Let A be a subset of IR. We dene
provided the right-hand side is not of the form 1 ; 1. If both IR f + d 0 and IR f ; d 0 are nite R (or equivalently, IR jf j d 0 < 1, since jf j = f + + f ; ), we say that f is integrable.
IR
fd 0=
IR
f+ d
0;;
IR
f d
;
A
where
fd 0=
I l A (x) =
is the indicator function of A.
IR
I l Af d
1 0
if x 2 A if x 2 A =
The Lebesgue integralRjust dened is related to the Riemann integral in one very important way: if R b the Riemann integral a f (x)dx is dened, then the Lebesgue integral a b] f d 0 agrees with the Riemann integral. The Lebesgue integral has two important advantages over the Riemann integral. The rst is that the Lebesgue integral is dened for more functions, as we show in the following examples. Example 1.5 Let Q be the set of rational numbers in 0 1], and consider f set, Q has Lebesgue measure zero, and so the Lebesgue integral of f over
= l Q . Being a countable I 0 1] is
1 To compute the Riemann integral 0 f (x)dx, we choose partition points 0 = x 0 < x1 < < xn = 1 and divide the interval 0 1] into subintervals x 0 x1] x1 x2] : : : xn;1 xn]. In each subinterval xk;1 xk ] there is a rational point q k , where f (qk ) = 1, and there is also an irrational point rk , where f (rk ) = 0. We approximate the Riemann integral from above by the upper sum
0 1]
f d 0 = 0:
n X
k=1 n X k=1
f (qk )(xk ; xk 1 ) =
;
n X
k=1 n X k=1
1 (xk ; xk 1 ) = 1
;
f (rk )(xk ; xk 1 ) =
;
0 (xk ; xk 1 ) = 0:
;
28 No matter how ne we take the partition of 0 1], the upper sum is always 1 and the lower sum is always 0. Since these two do not converge to a common value as the partition becomes ner, the Riemann integral is not dened. Example 1.6 Consider the function
f (x) =
if x = 0 if x 6= 0:
This is not a simple function because simple function cannot take the value function which lies between 0 and f is of the form
1.
Every simple
h(x) =
for some y
if x = 0 if x 6= 0
h d 0 = y 0 f0 g = 0 :
It follows that
IR
f d 0 = sup
Z
IR
hd
1 Now consider the Riemann integral ;1 f (x) dx, which for this function f is the same as the R 1 f (x) dx. When we partition ;1 1] into subintervals, one of these will contain Riemann integral ;1 R1 the point 0, and when we compute the upper approximating sum for ;1 f (x) dx, this point will contribute 1 times the length of the subinterval containing it. Thus the upper approximating sum is 1. On the other hand, the lower approximating sum is 0, and again the Riemann integral does not exist.
The Lebesgue integral has all linearity and comparison properties one would expect of an integral. In particular, for any two functions f and g and any real constant c,
IR
and whenever f (x)
(f + g ) d
0 0
IR
cf d
= c
IR Z
fd 0+ fd
0
IR
gd
IR
IR
fd
Z
IR
gd d 0:
A B
fd 0=
Z
A
fd 0+
Z
B
f d 0:
29
There are three convergence theorems satised by the Lebesgue integral. In each of these the situation is that there is a sequence of functions f n n = 1 2 : : : converging pointwise to a limiting function f . Pointwise convergence just means that
There are no such theorems for the Riemann integral, because the Riemann integral of the limiting function f is too often not dened. Before we state the theorems, we given two examples of pointwise convergence which arise in probability theory. Example 1.7 Consider a sequence of normal densities, each with variance mean n: 1 ; (x;n)2
fn (x) = p e
2
Example 1.8 Consider a sequence of normal densities, each with mean 0 and the n-th having vari1 ance n : r 2 2 fn (x) = n e; xn :
f (x) =
We have again IR fn d 0 = 1 for every n, so limn!1 IR fn d 0 = 1, but IR f d 0 = 0. The function f is not the Dirac delta; the Lebesgue integral of this function was already seen in Example 1.6 to be zero. Theorem 3.1 (Fatous Lemma) Let fn verging pointwise to a function f . Then
if x = 0 if x 6= 0:
IR
fd fd
lim inf n
!1
Z
IR
fn d 0 :
IR
nlim
!1
IR
fn d 0:
nlim IR fn d 0 = 1
!1
30 while IR f d 0 = 0. We could modify either Example 1.7 or 1.8 Rby setting g n = fn if n is even, R f but gn = 2R n if n is odd. Now IR gn d 0 = 1 if n is even, but IR gn d 0 = 2 if n is odd. The sequence f IR gn d 0 g1 has two cluster points, 1 and 2. R By denition, the smaller one, 1, is n=1 R lim inf n!1 IR gn d 0 and the larger one, 2, is lim supn!1 IR gn d 0 . Fatous Lemma guarantees that even the smaller cluster point will be greater than or equal to the integral of the limiting function. The key assumption in Fatous Lemma is that all the functions take only nonnegative values. Fatous Lemma does not assume much but it is is not very satisfying because it does not conclude that
IR
f d 0 = nlim
!1
IR
fn d 0: n = 1 2 : : : be a sequence of functions
for every x 2 IR:
There are two sets of assumptions which permit this stronger conclusion. Theorem 3.2 (Monotone Convergence Theorem) Let fn converging pointwise to a function f . Assume that
Z
IR
f d 0 = nlim
Z
IR
!1
fn d
Theorem 3.3 (Dominated Convergence Theorem) Let fn n = 1 2 : : : be a sequence of functions, which may take either positive or negative values, converging pointwise to a function f . Assume R that there is a nonnegative integrable function g (i.e., IR g d 0 < 1) such that
Then
f d 0 = nlim
!1
IR
fn d
, a nonempty set, called the sample space, which contains all possible outcomes of some random experiment;
F, a
-algebra of subsets of ;
IP , a probability measure on ( F ), i.e., a function which assigns to each set A 2 F a number IP (A) 2 0 1], which represents the probability that the outcome of the random experiment lies in the set A.
31
Remark 1.1 We recall from Homework Problem 1.4 that a probability measure IP has the following properties: (a)
IP ( ) = 0.
k=1
Ak =
! X
1
k=1
IP (Ak ):
IP (A1
(d) If A and B are sets in F and A
k=1
1
Ak = nlim IP (An ):
!1
, then
\ !
, then
k=1
Ak = nlim IP (An ):
!1
We have already seen some examples of nite probability spaces. We repeat these and give some examples of innite probability spaces as well. Example 1.9 Finite coin toss space. Toss a coin n times, so that is the set of all sequences of H and T which have n components. We will use this space quite a bit, and so give it a name: n . Let F be the collection of all subsets of n . Suppose the probability of H on each toss is p, a number between zero and one. Then the probability of T is q = 1 ; p. For each ! = (!1 !2 : : : !n ) in n , we dene For each A 2 F , we dene
X
! A
2
IP f!g:
(4.1)
We can dene IP (A) this way because A has only nitely many elements, and so only nitely many terms appear in the sum on the right-hand side of (4.1).
32 Example 1.10 Innite coin toss space. Toss a coin repeatedly without stopping, so that is the set of all nonterminating sequences of H and T . We call this space 1 . This is an uncountably innite space, and we need to exercise some care in the construction of the -algebra we will use here.
For each positive integer n, we dene Fn to be the -algebra determined by the rst n tosses. For example, F2 contains four basic sets,
AHH = f! = (!1 !2 !3 : : : ) !1 = H !2 = H g = The set of all sequences which begin with HH AHT = f! = (!1 !2 !3 : : : ) !1 = H !2 = T g = The set of all sequences which begin with HT ATH = f! = (!1 !2 !3 : : : ) !1 = T !2 = H g = The set of all sequences which begin with TH ATT = f! = (!1 !2 !3 : : : ) !1 = T !2 = T g = The set of all sequences which begin with TT:
In the -algebra F , we put every set in every -algebra Fn , where n ranges over the positive integers. We also put in every other set which is required to make F be a -algebra. For example, the set containing the single sequence Because sets.
F2 is a
fH on every tossg =
is also in F .
\
1
n=1
We next construct the probability measure IP on ( 1 F ) which corresponds to probability p 2 0 1] for H and probability q = 1 ; p for T . Let A 2 F be given. If there is a positive integer n such that A 2 Fn , then the description of A depends on only the rst n tosses, and it is clear how to dene IP (A). For example, suppose A = AHH ATH , where these sets were dened earlier. Then A is in F2. We set IP (AHH ) = p2 and IP (ATH ) = qp, and then we have
IP (A) = IP (AHH ATH ) = p2 + qp = (p + q )p = p: In other words, the probability of a H on the second toss is p.
33
Let us now consider a set A 2 F for which there is no positive integer n such that A 2 F . Such is the case for the set fH on every tossg. To determine the probability of these sets, we write them in terms of sets which are in Fn for positive integers n, and then use the properties of probability measures listed in Remark 1.1. For example,
and
\
1
n=1
A similar argument shows that if 0 < p < 1 so that 0 < q < 1, then every set in 1 which contains only one element (nonterminating sequence of H and T ) has probability zero, and hence very set which contains countably many elements also has probabiliy zero. We are in a case very similar to Lebesgue measure: every point has measure zero, but sets can have positive measure. Of course, the only sets which can have positive probabilty in 1 are those which contain uncountably many elements. In the innite coin toss space, we dene a sequence of random variables Y1
Yk (!) =
and we also dene the random variable
Y2 : : : by
1 0
if !k if !k
=H =T
X (! ) =
n X Y k (! ) k=1
2k
Since each Yk is either zero or one, X takes values in the interval 0 1]. Indeed, X (TTTT ) = 0, X (HHHH ) = 1 and the other values of X lie in between. We dene a dyadic rational m number to be a number of the form 2k , where k and m are integers. For example, 3 is a dyadic 4 rational. Every dyadic rational in (0,1) corresponds to two sequences ! 2 1 . For example,
X (HHTTTTT
) = X (HTHHHHH
) = 3: 4
The numbers in (0,1) which are not dyadic rationals correspond to a single ! have a unique binary expansion.
; these numbers
34 Whenever we place a probability measure IP on ( F ), we have a corresponding induced measure LX on 0 1]. For example, if we set p = q = 1 in the construction of this example, then we have 2
Continuing this process, we can verify that for any positive integers k and m satisfying
m;1 < m
2k 2k
we have
In other words, the LX -measure of all intervals in 0 1] whose endpoints are dyadic rationals is the same as the Lebesgue measure of these intervals. The only way this can be is for LX to be Lebesgue measure. It is interesing to consider what L X would look like if we take a value of p other than 1 when we 2 construct the probability measure IP on .
m LX m2; 1 2k = 21k : k
We conclude this example with another look at the Cantor set of Example 3.2. Let pairs be the subset of in which every even-numbered toss is the same as the odd-numbered toss immediately preceding it. For example, HHTTTTHH is the beginning of a sequence in pairs , but HT is not. Consider now the set of real numbers
C = fX ( ! ) ! 2
0
pairs g:
The numbers between ( 1 1 ) can be written as X (! ), but the sequence ! must begin with either 4 2 1 3 TH or HT . Therefore, none of these numbers is in C 0. Similarly, the numbers between ( 16 16 ) can be written as X (! ), but the sequence ! must begin with TTTH or TTHT , so none of these numbers is in C 0. Continuing this process, we see that C 0 will not contain any of the numbers which were removed in the construction of the Cantor set C in Example 3.2. In other words, C 0 C. With a bit more work, one can convince onself that in fact C 0 = C , i.e., by requiring consecutive coin tosses to be paired, we are removing exactly those points in 0 1] which were removed in the Cantor set construction of Example 3.2.
35
In addition to tossing a coin, another common random experiment is to pick a number, perhaps using a random number generator. Here are some probability spaces which correspond to different ways of picking a number at random. Example 1.11 Suppose we choose a number from IR in such a way that we are sure to get either 1, 4 or 16. Furthermore, we construct the experiment so that the probability of getting 1 is 4 , the probability of 9 getting 4 is 4 and the probability of getting 16 is 1 . We describe this random experiment by taking 9 9 to be IR, F to be B(IR), and setting up the probability measure so that
5]
The probability measure described in this example is L S2 , the measure induced by the stock price S2 , when the initial stock price S 0 = 4 and the probability of H is 1 . This distribution was discussed 3 immediately following Denition 2.8. Example 1.12 Uniform distribution on 0 1]. Let = 0 1] and let F = B( 0 1]), the collection of all Borel subsets containined in 0 1]. For each Borel set A 0 1], we dene IP (A) = 0 (A) to be the Lebesgue measure of the set. Because 0 1] = 1, this gives us a probability measure. 0
This probability space corresponds to the random experiment of choosing a number from 0 1] so that every number is equally likely to be chosen. Since there are innitely mean numbers in 0 1], this requires that every number have probabilty zero of being chosen. Nonetheless, we can speak of the probability that the number chosen lies in a particular set, and if the set has uncountably many points, then this probability can be positive. I know of no way to design a physical experiment which corresponds to choosing a number at random from 0 1] so that each number is equally likely to be chosen, just as I know of no way to toss a coin innitely many times. Nonetheless, both Examples 1.10 and 1.12 provide probability spaces which are often useful approximations to reality. Example 1.13 Standard normal distribution. Dene the standard normal density
'(x) = p1 e 2 : 2
;
x2
Let
IP (A) =
' d 0:
(4.2)
36 If A in (4.2) is an interval
a b], then we can write (4.2) as the less mysterious Riemann integral: Z b 1 x2 IP a b] = p e 2 dx:
;
This corresponds to choosing a point at random on the real line, and every single point has probability zero of being chosen, but if a set A is given, then the probability the point is in that set is given by (4.2). The construction of the integral in a general probability space follows the same steps as the construction of Lebesgue integral. We repeat this construction below. Denition 1.14 Let ( F IP ) be a probability space, and let X be a random variable on this space, i.e., a mapping from to IR, possibly also taking the values 1. If X is an indicator, i.e,
X (! ) = l A (!) = I
for some set A 2 F , we dene If X is a simple function, i.e,
1 0
if ! if !
2A 2 Ac
X dIP = IP (A):
n X k=1
X (! ) =
n X Z
ck l Ak (!) I
n X
X dIP =
k=1
ck
I l Ak dIP =
k=1
ck IP (Ak ):
X dIP
= sup
Y dIP Y
is simple and Y (! )
In fact, we can always construct a sequence of simple functions Yn and Y (! ) = limn!1 Yn (! ) for every !
n = 1 2 : : : such that
0 Y1 (! ) Y2 (! ) Y3 (! ) : : : for every ! 2
X dIP = nlim
!1
Yn dIP:
37
X + dIP < 1
Z
;
X dIP < 1
;
X dIP =
X + dIP
; ; X dIP:
;
X dIP = IEX =
I l A X dIP: X dIP:
The above integral has all the linearity and comparison properties one would expect. In particular, if X and Y are random variables and c is a real constant, then
(X + Y ) dIP =
cX dIP = c X dP
, then
X dIP +
Y dIP
If X (! )
X dIP
Y dIP:
In fact, we dont need to have X (! ) Y (! ) for every ! 2 in order to reach this conclusion; it is enough if the set of ! for which X (! ) Y (! ) has probability one. When a condition holds with probability one, we say it holds almost surely. Finally, if A and B are disjoint subsets of and X is a random variable, then
A B
X dIP =
X dIP +
X dIP:
We restate the Lebesgue integral convergence theorem in this more general context. We acknowledge in these statements that conditions dont need to hold for every ! ; almost surely is enough. Theorem 4.4 (Fatous Lemma) Let Xn n = 1 2 : : : be a sequence of almost surely nonnegative random variables converging almost surely to a random variable X . Then
X dIP
lim inf n
!1 !1
Xn dIP
or equivalently,
38 Theorem 4.5 (Monotone Convergence Theorem) Let Xn n = 1 2 : : : be a sequence of random variables converging almost surely to a random variable X . Assume that
0 X1 X2 X3
Then
almost surely:
X dIP = nlim
!1
Xn dIP
or equivalently,
Theorem 4.6 (Dominated Convergence Theorem) Let Xn n = 1 2 : : : be a sequence of random variables, converging almost surely to a random variable X . Assume that there exists a random variable Y such that jXnj Y almost surely for every n: Then
X dIP = nlim
!1
Xn dIP
or equivalently,
In Example 1.13, we constructed a probability measure on (IR B(IR)) by integrating the standard R normal density. In fact, whenever ' is a nonnegative function dened on R satisfying IR ' d 0 = 1, we call ' a density and we can dene an associated probability measure by
IP (A) =
'd
(4.3)
We shall often have a situation in which two measure are related by an equation like (4.3). In fact, the market measure and the risk-neutral measures in nancial markets are related this way. We say that ' in (4.3) is the Radon-Nikodym derivative of dIP with respect to 0 , and we write
dIP '= d :
0
(4.4)
The probability measure IP weights different parts of the real line according to the density '. Now suppose f is a function on (R B(IR) IP ). Denition 1.14 gives us a value for the abstract integral
Z Z
IR IR
f dIP:
0
f' d
which is an integral with respec to Lebesgue measure over the real line. We want to show that
IR
f dIP =
IR
f' d
(4.5)
39
dIP an equation which is suggested by the notation introduced in (4.4) (substitute d 0 for ' in (4.5) and cancel the d 0 ). We include a proof of this because it allows us to illustrate the concept of the standard machine explained in Williamss book in Section 5.12, page 5.
The standard machine argument proceeds in four steps. Step 1. Assume that f is an indicator function, i.e., f (x) that case, (4.5) becomes Z
IR. In
IP (A) = ' d 0: A This is true because it is the denition of IP (A). Step 2. Now that we know that (4.5) holds when f is an indicator function, assume that f
simple function, i.e., a linear combination of indicator functions. In other words,
is a
f (x) =
n X k=1
ck hk (x)
IR
f dIP =
= = = =
Z "X n
f' d 0:
Step 3. Now that we know that (4.5) holds when f is a simple function, we consider a general nonnegative function f . We can always construct a sequence of nonnegative simple functions fn n = 1 2 : : : such that
We let n ! 1 and use the Monotone Convergence Theorem on both sides of this equality to get Z Z
IR
fn dIP =
IR
fn ' d
0 for every n:
IR
f dIP =
IR
f' d 0:
40 Step 4. In the last step, we consider an integrable function f , which can take both positive and negative values. By integrable, we mean that
IR
From Step 3, we have
f + dIP
<1
IR
f dIP < 1:
;
Z ZIR
IR
f + dIP = f dIP =
;
Z ZIR
IR
f +' d
;
f ' d 0:
IR
f dIP =
= =
ZIR ZIR
R
0;
IR Z
f dIP
;
IR
f 'd
;
1.5 Independence
In this section, we dene and discuss the notion of independence in a general probability space ( F IP ), although most of the examples we give will be for coin toss space.
1.5.1
Independence of sets
IP (A \ B) = IP (A)IP (B):
2 F are independent if
Suppose a random experiment is conducted, and ! is the outcome. The probability that ! 2 A is IP (A). Suppose you are not told !, but you are told that ! 2 B . Conditional on this information, the probability that ! 2 A is The sets A and B are independent if and only if this conditional probability is the uncondidtional probability IP (A), i.e., knowing that ! 2 B does not change the probability you assign to A. This discussion is symmetric with respect to A and B ; if A and B are independent and you know that ! 2 A, the conditional probability you assign to B is still the unconditional probability IP (B ).
(A \ IP (AjB) = IP IP (B)B) :
Whether two sets are independent depends on the probability measure IP . For example, suppose we toss a coin twice, with probability p for H and probability q = 1 ; p for T on each toss. To avoid trivialities, we assume that 0 < p < 1. Then
(5.1)
41
TH g. In words, A is the set H on the rst toss and B is the = fHT g. We compute IP (A) = p2 + pq = p IP (B ) = 2pq IP (A)IP (B ) = 2p2q IP (A \ B) = pq:
If p = 1 , then IP (B ), the probability of one head and one tail, is 1 . If you are told that the coin 2 2 tosses resulted in a head on the rst toss, the probability of B , which is now the probability of a T on the second toss, is still 1 . 2 Suppose however that p = 0:01. By far the most likely outcome of the two coin tosses is TT , and the probability of one head and one tail is quite small; in fact, IP (B ) = 0:0198. However, if you are told that the rst toss resulted in H , it becomes very likely that the two tosses result in one head and one tail. In fact, conditioned on getting a H on the rst toss, the probability of one H and one T is the probability of a T on the second toss, which is 0:99.
1 2.
1.5.2
Independence of -algebras
Denition 1.16 Let G and H be sub- -algebras of F . We say that G and H are independent if every set in G is independent of every set in H, i.e,
G = F1 be the
-algebra
fHH HT g fTH TT g:
Let H be the -albegra determined by the second toss: H contains the sets
fHH TH g fHT TT g:
These two -algebras are independent. For example, if we choose the set fHH the set fHH TH g from H, then we have
HT g from G and
42 Example 1.14 illustrates the general principle that when the probability for a sequence of tosses is dened to be the product of the probabilities for the individual tosses of the sequence, then every set depending on a particular toss will be independent of every set depending on a different toss. We say that the different tosses are independent when we construct probabilities this way. It is also possible to construct probabilities such that the different tosses are not independent, as shown by the following example.
= fHH HT TH TT g to be 1 1 IP fHH g = 9 IP fHT g = 2 IP fTH g = 1 IP fTT g = 3 9 3 and for every set A , dene IP (A) to be the sum of the probabilities of the elements in A. Then IP ( ) = 1, so IP is a probability measure. Note that the sets fH on rst tossg = fHH HT g and fH on second tossg = fHH TH g have probabilities IP fHH HT g = 1 and IP fHH TH g = 3
Example 1.15 Dene IP for the individual elements of independent.
4 4 9 , so the product of the probabilities is 27 . On the other hand, the intersection of fHH HT g and fHH TH g contains the single element fHH g, which has probability 1 . These sets are not 9
1.5.3
Denition 1.17 We say that two random variables X and Y are independent if the -algebras they generate (X ) and (Y ) are independent. In the probability space of three independent coin tosses, the price S 2 of the stock at time 2 is independent of S3 . This is because S2 depends on only the rst two coin tosses, whereas S3 is S2 S2 either u or d, depending on whether the third coin toss is H or T .
Denition 1.17 says that for independent random variables X and Y , every set dened in terms of X is independent of every set dened in terms of Y . In the case of S2 and S3 just considered, for exS2 ample, the sets fS2 = are indepedent sets.
n S3
Suppose X and Y are independent random variables. We dened earlier the measure induced by X on IR to be Similarly, the measure induced by Y is Now the pair (X pair
Y ) takes values in the plane IR 2, and we can dene the measure induced by the
The set C in this last equation is a subset of the plane IR 2 . In particular, C could be a rectangle, i.e, a set of the form A B , where A IR and B IR. In this case,
LX Y (C ) = IP f(X Y ) 2 C g C IR2:
f(X Y ) 2 A Bg = fX 2 Ag \ fY 2 Bg
43
LX Y (A B) = IP fX 2 Ag \ fY 2 Bg = IP fX 2 AgIP fY 2 B g = LX (A)LY (B ):
(5.2)
In other words, for independent random variables X and Y , the joint distribution represented by the measure LX Y factors into the product of the marginal distributions represented by the measures LX and LY . A joint density for (X
LX Y (A B) = Z
Z Z
Not every pair of random variables (X Y ) has a joint density, but if a pair does, then the random variables X and Y have marginal densities dened by
A B
fX Y (x y) dx dy:
fX (x) =
These have the properties
;1
fX Y (x ) d
fY (y)
;1
fX Y ( y ) d :
LX (A) = LY (B) =
Z ZA
B
Suppose X and Y have a joint density. Then X and Y are independent variables if and only if the joint density is the product of the marginal densities. This follows from the fact that (5.2) is equivalent to independence of X and Y . Take A = (;1 x] and B = (;1 y ], write (5.1) in terms of densities, and differentiate with respect to both x and y . Theorem 5.7 Suppose X and Y are independent random variables. Let g and h be functions from IR to IR. Then g(X ) and h(Y ) are also independent random variables. P ROOF : Let us denote W = g (X ) and Z a typical set in (W ) is of the form
44 Denition 1.18 Let X1 X2 : : : be a sequence of random variables. We say that these random variables are independent if for every sequence of sets A1 2 (X1) A2 2 (X2) : : : and for every positive integer n,
IP (A1 \ A2 \
IP (An ):
1.5.4
Theorem 5.8 If two random variables X and Y are independent, and if g and h are functions from IR to IR, then
IP fX 2 Ag \ fY 2 Bg = IP fX 2 AgIP fY 2 Bg
which is true because X and Y are independent. Now use the standard machine to get the result for general functions g and h. The variance of a random variable X is dened to be Var(X ) = IE
X ; IEX ]2:
i
X and Y
According to Theorem 5.8, for independent random variables, the covariance is zero. If both have positive variances, we dene their correlation coefcient
X and Y
45
IP fY
Being standard normal, both X and Y have expected value zero. Therefore, Cov(X
yg = IP fY = IP fX = IP fX 1 = 2 IP fX
2 IR, we have y and Z = 1g + IP fY y and Z = ;1g y and Z = 1g + IP f;X y and Z = ;1g y gIP fZ = 1g + IP f;X y gIP fZ = ;1g y g + 1 IP f;X y g: 2
Y ) = IE XY ] = IE X 2Z ] = IEX 2 IEZ = 1 0 = 0: Y )?
Where in IR2 does the measure LX Y put its mass, i.e., what is the distribution of (X
We conclude this section with the observation that for independent random variables, the variance of their sum is the sum of their variances. Indeed, if X and Y are independent and Z = X + Y , then Var(Z )
= IE (Z ; IEZ )2 i = IE X + Y ; IEX ; IEY )2 h i = IE (X ; IEX )2 + 2(X ; IEX )(Y ; IEY ) + (Y ; IEY )2 = Var(X ) + 2IE X ; IEX ]IE Y ; IEY ] + Var(Y ) = Var(X ) + Var(Y ):
This argument extends to any nite number of random variables. If we are given independent random variables X1 X2 : : : Xn , then Var(X1 + X2 +
+ Xn ) = Var(X1) + Var(X2) +
+ Var(Xn ):
(5.3)
1.5.5
We now return to property (k) for conditional expectations, presented in the lecture dated October 19, 1995. The property as stated there is taken from Williamss book, page 88; we shall need only the second assertion of the property: (k) If a random variable X is independent of a -algebra H, then
IE X jH] = IEX:
The point of this statement is that if X is independent of H, then the best estimate of X based on the information in H is IEX , the same as the best estimate of X based on no information.
46 To show this equality, we observe rst that IEX is H-measurable, since it is not random. We must also check the partial averaging property
If X is an indicator of some set B , which by assumption must be independent of H, then the partial averaging equation we must check is
IEX dIP =
The left-hand side of this equation is IP (A)IP (B ), and the right hand side is
IP (B) dIP =
I l B dIP:
I I l Al B dIP =
I l A B dIP = IP (A \ B):
\
The partial averaging equation holds because A and B are independent. The partial averaging equation for general X independent of H follows by the standard machine.
1.5.6
There are two fundamental theorems about sequences of independent random variables. Here is the rst one. Theorem 5.9 (Law of Large Numbers) Let X1 X2 : : : be a sequence of independent, identically distributed random variables, each with expected value and variance 2. Dene the sequence of averages Then Yn converges to
Yn = X1 + X2 + + Xn n = 1 2 : : :: n almost surely as n ! 1.
We are not going to give the proof of this theorem, but here is an argument which makes it plausible. We will use this argument later when developing stochastic calculus. The argument proceeds in two steps. We rst check that IEYn = for every n. We next check that Var(Yn ) ! 0 as n ! 0. In other words, the random variables Yn are increasingly tightly distributed around as n ! 1. For the rst step, we simply compute
1 + IEXn] = n | + + + } = : ] {z
n times
For the second step, we rst recall from (5.3) that the variance of the sum of independent random variables is the sum of their variances. Therefore, Var(Yn ) = As n ! 1, we have Var(Yn ) ! 0.
n X k=1
Var
n Xk = X 2 = 2 : 2 n n k=1 n
47
1.5.7
The Law of Large Numbers is a bit boring because the limit is nonrandom. This is because the denominator in the denition of Y n is so large that the variance of Yn converges to zero. If we want p to prevent this, we should divide by n rather than n. In particular, if we again have a sequence of independent, identically distributed random variables, each with expected value and variance 2 , but now we set then each Zn has expected value zero and Var(Zn ) =
+ (Xn ; )
2
As n ! 1, the distributions of all the random variables Z n have the same degree of tightness, as measured by their variance, around their expected value 0. The Central Limit Theorem asserts that as n ! 1, the distribution of Z n approaches that of a normal random variable with mean (expected value) zero and variance 2. In other words, for every set A IR,
k=1
Xk ; pn
n X k=1
2 n = :
1 Z e lim IP fZn 2 Ag = p n 2 A
!1
x2 2 2
dx:
48
Chapter 2
Conditional Expectation
Please see Hulls book (Section 9.6.)
Note that we are not specifying the probability of heads here. Consider a sequence of 3 tosses of the coin (See Fig. 2.1) The collection of all possible outcomes (i.e. sequences of tosses of length 3) is
Each Sk is a random variable dened on the set . More precisely, let F = P ( ). Then F is a -algebra and ( F ) is a measurable space. Each Sk is an F -measurable function !IR, that is, ; Sk 1 is a function B!F where B is the Borel -algebra on I . We will see later that Sk is in fact R 49
50
= 3 S2 (HH) =u 2S0 2 = S (H) = uS0 1 1= S0 1 = S (T) = dS0 1 2 = S2 (TT) = d 2S0 3 = S3 (TTT) = d 3 S0 = 3 2 = 2 = S2 (HT) = ud S 0 S2 (TH) = ud S 0 = 3 3 = 3 = S3 (HHT) = u2 d S0 S3 (HTH) = u2 d S0 S3 (THH) = u2 d S0 3 S3 (HHH) = u S 0
Figure 2.1: A three coin period binomial model. measurable under a sub- -algebra of F . Recall that the Borel -algebra B is the -algebra generated by the open intervals of I . In this course we will always deal with subsets of I that belong to B. R R For any random variable X dened on a sample space
4
and any y
fX yg = f! 2 X (!) yg: The sets fX < y g fX y g fX = y g etc, are dened similarly. Similarly for any subset B of IR,
we dene
fX 2 Bg = f! 2 X (!) 2 Bg:
4
Assumption 2.1
u > d > 0.
2.2 Information
Denition 2.1 (Sets determined by the rst k tosses.) We say that a set A is determined by the rst k coin tosses if, knowing only the outcome of the rst k tosses, we can decide whether the outcome of all tosses is in A. In general we denote the collection of sets determined by the rst k tosses by F k . It is easy to check that F k is a -algebra. Note that the random variable Sk is F k -measurable, for each k = 1
2 : : : n.
Example 2.1 In the 3 coin-toss example, the collection F 1 of sets determined by the rst toss consists of:
51
The collection F 2 of sets determined by the rst two tosses consists of: 1. 2. 3.
4. 5. The complements of the above sets, 6. Any union of the above sets (including the complements), 7. and .
Denition 2.2 (Information carried by a random variable.) Let X be a random variable !IR. We say that a set A is determined by the random variable X if, knowing only the value X (!) of the random variable, we can decide whether or not ! 2 A. Another way of saying this is that for every y 2 IR, either X ;1(y ) A or X ;1 (y ) \ A = . The collection of susbets of determined by X is a -algebra, which we call the -algebra generated by X , and denote by (X ). If the random variable X takes nitely many different values, then tion of sets
fX 1(X (!))j! 2 g
;
these sets are called the atoms of the -algebra In general, if X is a random variable
(X ) = fX 1(B ) B 2 Bg:
Example 2.2 (Sets determined by S2 ) The -algebra generated by S2 consists of the following sets: 1. 2. 3.
AHH = fHHH HHT g = f! 2 S2 (!) = u2S0 g, AT T = fTTH T TT g = fS2 = d2S0 g AHT AT H = fS2 = udS0g
4. Complements of the above sets, 5. Any union of the above sets, 6. = fS2 (!) 2 g, 7. = fS2 (!) 2 IRg.
52
the coin tosses are independent, so that, e.g., IP (HHT ) = p2q etc.
IP (A) = P! A IP (!), 8A
4 2
IEX =
4
X
!
2
X (!)IP (! ):
1 0
if ! if !
If A
then
IA (!) =
4
and
IE (IAX ) =
Z
A
XdIP =
2A 62 A X
2
X (!)IP (! ):
2.3.1
An example
Let us estimate S1, given S2. Denote the estimate by IE (S1jS2). From elementary probability, IE (S1jS2) is a random variable Y whose value at ! is dened by
IE (S1jS2) should depend on !, i.e., it is a random variable. If the value of S2 is known, then the value of IE (S 1jS2 ) should also be known. In particular, If ! = HHH or ! = HHT , then S2 (! ) = u2 S0. If we know that S2(! ) = u2 S0 , then even without knowing ! , we know that S 1 (! ) = uS0. We dene IE (S1jS2)(HHH ) = IE (S1jS2)(HHT ) = uS0: If ! = TTT or ! = TTH , then S2(! ) = d2S0 . If we know that S2(! ) = d2S0 , then even without knowing ! , we know that S 1 (! ) = dS0. We dene IE (S1jS2)(TTT ) = IE (S1jS2 )(TTH ) = dS0 :
53
udS0, then we do not know whether S1 = uS0 or S1 = dS0. We then take a weighted
average:
Furthermore,
Z
A
For !
2 A we dene
Then
IE (S1jS2)dIP =
Z
A
S1dIP:
IE (S1jS2)(!) = g (S2(!))
if x = u2 S0 if x = udS0 if x = d2 S0
IE (S1jS2 = x) = g (x)
where g is the function dened above. The random variable IE (S1jS2) has two fundamental properties:
IE (S1jS2)dIP =
Z
A
S1dIP:
F IP ) be a probability space, and let G be a sub- -algebra of F . Let X be a random variable F IP ). Then IE (X jG) is dened to be any random variable Y that satises:
Y is G -measurable,
Y dIP =
XdIP:
Uniqueness. There can be more than one random variable Y satisfying the above properties, but if Y 0 is another one, then Y = Y 0 almost surely, i.e., IP f! 2 Y (!) = Y 0 (!)g = 1: Notation 2.1 For random variables X
Existence. There is always a random variable Y satisfying the above properties (provided that IE jX j < 1), i.e., conditional expectations always exist.
IE (X jY ) = IE (X j (Y )):
Here are some useful ways to think about IE (X jG): A random experiment is performed, i.e., an element ! of is selected. The value of ! is partially but not fully revealed to us, and thus we cannot compute the exact value of X (!). Based on what we know about ! , we compute an estimate of X (!). Because this estimate depends on the partial information we have about ! , it depends on !, i.e., IE X jY ](!) is a function of ! , although the dependence on ! is often not shown explicitly. If the -algebra G contains nitely many sets, there will be a smallest set A in G containing ! , which is the intersection of all sets in G containing ! . The way ! is partially revealed to us is that we are told it is in A, but not told which element of A it is. We then dene IE X jY ](!) to be the average (with respect to IP ) value of X over this set A. Thus, for all ! in this set A, IE X jY ](!) will be the same.
2.3.3
A
We can rewrite this as
IE (X jG)dIP =
Z
A
XdIP 8A 2 G :
(3.1)
(3.2)
(3.3)
55
Proof: To see this, rst use (3.2) and linearity of expectations to prove (3.3) when V is a simple G -measurable random variable, i.e., V is of the form V = Pn=1 ck IAK , where each Ak is in G and k each ck is constant. Next consider the case that V is a nonnegative G -measurable random variable, but is not necessarily simple. Such a V can be written as the limit of an increasing sequence of simple random variables Vn ; we write (3.3) for each Vn and then pass to the limit, using the Monotone Convergence Theorem (See Williams), to obtain (3.3) for V . Finally, the general G measurable random variable V can be written as the difference of two nonnegative random-variables V = V + ; V ; , and since (3.3) holds for V + and V ; it must hold for V as well. Williams calls this argument the standard machine (p. 56). Based on this lemma, we can replace the second condition in the denition of a conditional expectation (Section 2.3.2) by: (b) For every G -measurable random-variable V , we have
(3.4)
Please see Willams p. 88. Proof sketches of some of the properties are provided below. (a)
(b)
IE (IE (X jG)) = IE (X ): Proof: Just take A in the partial averaging property to be . The conditional expectation of X is thus an unbiased estimator of the random variable X . If X is G -measurable, then IE (X jG) = X: Proof: The partial averaging property holds trivially when Y is replaced by X . And since X is G -measurable, X satises the requirement (a) of a conditional expectation as well. If the information content of G is sufcient to determine X , then the best estimate of X based on G is X itself. IE (a1X1 + a2X2jG) = a1IE (X1jG) + a2IE (X2jG):
0 almost surely, then
IE (X jG) 0:
Proof: Take A = f! 2 R (X jG)(! ) < 0g. This set is in G since IE (X jG) is G -measurable. IE R Partial averaging implies A IE (X jG)dIP = A XdIP . The right-hand side is greater than or equal to zero, and the left-hand side is strictly negative, unless IP (A) = 0. Therefore, IP (A) = 0.
IE ( (X )jG)
Recall the usual Jensens Inequality: IE
(X )
H is a sub-
8A 2 G .
Take Y = Z:IE (X jG). Then Y satises (a) (a product of G -measurable random variables is G-measurable). Y also satises property (b), as we can check below:
( (X ) G), then
IE (X j (G H)) = IE (X jG):
In particular, if X is independent of H, then
IE (X jH) = IE (X ):
If H is independent of X and G , then nothing is gained by including the information content of H in the estimation of X .
57
2.3.5
Recall that F 1
= f AH AT g. Notice that IE (S2jF 1 ) must be constant on AH and AT . Now since IE (S2jF 1 ) must satisfy the partial averaging property,
Z
We compute
AH AT
Z Z
AH AT
S2 dIP S2dIP:
Z
AH
Z
AH
IE (S2jF 1 )(!) = pu2 S0 + qudS0 8! 2 AH : IE (S2jF 1)(! ) = pu2 S0 + qudS0 = (pu + qd)uS0 = (pu + qd)S1(!) 8! 2 AH
Similarly,
IE IE (S3jF 2)jF 1] = IE (pu + qd)S2jF 2] = (pu + qd)IE (S2jF 1) = (pu + qd)2S1: This nal expression is IE (S 3jF 1).
(linearity)
58
2.4 Martingales
The ingredients are: A probability space ( A sequence of
F . Such a sequence of
Conditions for a martingale:
1. Each Mk is F k -measurable. If you know the information in F k , then you know the value of Mk . We say that the process fMk g is adapted to the ltration fF k g. 2. For each k, IE (Mk+1 jF k ) = Mk . Martingales tend to go neither up nor down. A supermartingale tends to go down, i.e. the second condition above is replaced by IE (M k+1 jF k ) Mk ; a submartingale tends to go up, i.e. IE (M k+1jF k ) Mk .
Example 2.3 (Example from the binomial model.) For k
IE(S1 jF 0 )dIP =
S1 dIP:
In conclusion, If (pu + qd) = 1 then fSk If (pu + qd) 1 then fSk If (pu + qd) 1 then fSk
Chapter 3
Arbitrage Pricing
3.1 Binomial Pricing
Return to the binomial pricing model Please see: Cox, Ross and Rubinstein, J. Financial Economics, 7(1979), 229263, and Cox and Rubinstein (1985), Options Markets, Prentice-Hall.
Example 3.1 (Pricing a Call Option) Suppose u = 2 d = 0:5 r = 25%(interest rate), S0 = 50. (In this and all examples, the interest rate quoted is per unit time, and the stock prices S0 S1 : : : are indexed by the same time periods). We know that
S1 (!) =
100 25
if ! 1 if ! 1
=H =T
Find the value at time zero of a call option to buy one share of stock at time 1 for $50 (i.e. the strike price is $50). The value of the call at time 1 is
50 0
if ! 1 if ! 1
=H =T
Suppose the option sells for $20 at time 0. Let us construct a portfolio: 1. Sell 3 options for $20 each. Cash outlay is ;$60: 2. Buy 2 shares of stock for $50 each. Cash outlay is $100. 3. Borrow $40. Cash outlay is ;$40:
59
60
This portfolio thus requires no initial investment. For this portfolio, the cash outlay at time 1 is: Pay off option Sell stock Pay off debt
$0 The arbitrage pricing theory (APT) value of the option at time 0 is V0 = 20.
Assumptions underlying APT: Unlimited short selling of stock. Unlimited borrowing. No transaction costs. Agent is a small investor, i.e., his/her trading does not move the market. Important Observation: The APT value of the option does not depend on the probabilities of H and T .
;;;;; ;;;;;
!1 = T $0 ;$50 $50
negative).
0 shares of stock at time 0. ( 0 is also to be determined later) Invest V0 ; 0S0 in the money market, at risk-free interest rate r. (V0 ;
Buy Then wealth at time 1 is
0S0 might be
X1 =
4
0 so that
X1 = V1
regardless of whether the stock goes up or down.
61
The last condition above can be expressed by two equations (which is fortunate since there are two unknowns):
(1 + r)V0 + (1 + r)V0 +
(2.1) (2.2)
Note that this is where we use the fact that the derivative security value V k is a function of Sk , i.e., when Sk is known for a given ! , Vk is known (and therefore non-random) at that ! as well. Subtracting the second equation above from the rst gives
0=
Plug the formula (2.3) for
(2.3)
0 into (2.1):
We have already assumed u > d > 0. We now also assume d be an arbitrage opportunity). Dene
4 4
Then p > 0 and q > 0. Since p + q = 1, we have 0 < p < 1 and q = 1 ; p. Thus, p ~ ~ ~ ~ ~ ~ ~ ~ probabilities. We will return to this later. Thus the price of the call at time 0 is given by
r 1 p = 1 + ;; d q = u ; ; ; r : ~ ~ u d u d
1 ~ ~ V0 = 1 + r pV1(H ) + qV1(T )]:
q are like ~
(2.4)
f IP on
f f f IP is called the risk-neutral probability measure. We denote by IE the expectation under IP . Equation 2.4 says
f 1 V0 = IE 1 + r V1 :
62
F k gn=0 is a martingale. k
f IE (1 + r)
= =
3.3.1
Portfolio Process
=(
:::
n 1), where
;
3.3.2
Xk+1 =
=
Then each Xk is F k -measurable.
(3.1) (3.2)
f P Theorem 3.12 Under I , the discounted self-nancing portfolio process value f(1 + r) ;k Xk is a martingale.
Proof: We have
F k gn=0 k
63
f IE (1 + r) (k+1)Xk+1 jF k ] f = IE (1 + r) k Xk jF k ] f +IE (1 + r) (k+1) k Sk+1 jF k ] f ;IE (1 + r) k k Sk jF k ] = (1 + r) k Xk (requirement (b) of conditional exp.) f + k IE (1 + r) (k+1) Sk+1 jF k ] (taking out what is known) ;(1 + r) k k Sk (property (b)) = (1 + r) k Xk (Theorem 3.11)
; ; ; ; ; ; ; ;
Xm(!) = Vm(!) 8! 2 : In this case, for k = 0 1 : : : m, we call Xk the APT value at time k of V m .
Theorem 4.13 (Corollary to Theorem 3.12) If a simple European security Vm is hedgeable, then for each k = 0 1 : : : m, the APT value at time k of V m is
f Vk = (1 + r)k IE (1 + r) m VmjF k ]:
4 ;
(4.1)
= 0 1 : : : m ; 1, then we also have f IE Mm jF k ] = Mk k = 0 1 : : : m ; 1: (4.2) When k = m ; 1, the equation (4.2) follows directly from the martingale property. For k = m ; 2,
f ff IE Mm jF m 2 ] = IE IE Mm jF m 1 ]jF m 2 ] f = IE Mm 1 jF m 2 ] = Mm 2 :
; ; ; ; ; ;
64 We can continue by induction to obtain (4.2). If the simple European security Vm is hedgeable, then there is a portfolio process whose selfnancing value process X0 X1 : : : Xm satises Xm = Vm . By denition, Xk is the APT value at time k of Vm . Theorem 3.12 says that
Therefore,
f Xk = (1 + r)k IE (1 + r) m VmjF k ]:
;
(5.1)
k (!1
f E Starting with initial wealth V 0 = I (1 + r);m Vm ], the self-nancing value of the portfolio process 0 1 : : : m;1 is the process V0 V1 : : : Vm.
Proof: Let V0 : : : Vm;1 and 0 : : : m;1 be dened by (5.1) and (5.2). Set X0 = V0 and dene the self-nancing value of the portfolio process 0 : : : m;1 by the recursive formula 3.2:
(5.2)
Xk+1 =
We need to show that
k Sk+1 + (1 + r)(Xk ; k Sk ):
Xk = Vk
8k 2 f0 1 : : : mg:
(5.3)
We proceed by induction. For k = 0, (5.3) holds by denition of X 0. Assume that (5.3) holds for some value of k, i.e., for each xed (!1 : : : !k ), we have
Xk (!1 : : : !k ) = Vk (!1 : : : !k ):
65
f IE (1 + r)
(k+1) Vk+1 jF k ]
= = =
Since (!1 : : : !k ) will be xed for the rest of the proof, we simplify notation by suppressing these symbols. For example, we write the last equation as
We compute
Xk+1 (H ) = k Sk+1 (H ) + (1 + r)(Xk ; k Sk ) = k (Sk+1 (H ) ; (1 + r)Sk ) + (1 + r)Vk V (H ) ; V = Sk+1 (H ) ; Sk+1 (T ) (Sk+1 (H ) ; (1 + r)Sk ) k+1 k+1 (T ) +~Vk+1 (H ) + q Vk+1 (T ) p ~ (H ; Vk = Vk+1uS ) ; dS+1 (T ) (uSk ; (1 + r)Sk ) k k +~Vk+1 (H ) + q Vk+1 (T ) p ~ = (Vk+1 (H ) ; Vk+1 (T )) u ; 1 ; r + pVk+1 (H ) + qVk+1 (T ) ~ ~ u;d = (Vk+1 (H ) ; Vk+1 (T )) q + pVk+1 (H ) + q Vk+1 (T ) ~ ~ ~ = Vk+1 (H ):
66
Chapter 4
Example 4.1 (Lookback Option) u = 2 d = 0:5 r = 0:25 S0 = 4 p = 1+;;d = 0:5 q = 1 ; p = ~ ur d ~ ~ Consider a simple European derivative security with expiration 2, with payoff given by (See Fig. 4.1):
0:5:
V2 = 0max2(Sk ; 5)+ : k
Notice that
V2 (HH) = 11 V2 (HT) = 3 6= V2 (TH) = 0 V2 (T T ) = 0: 1 ~ V1 (H) = 1 + r pV2 (HH) + qV2 (HT)] = 4 0:5 11 + 0:5 3] = 5:60 ~ 5 V1 (T) = 4 0:5 0 + 0:5 0] = 0 5 4 0:5 5:60 + 0:5 0] = 2:24: V0 = 5
0
68
S2 (HH) = 16
S2 (TT) = 1
X1 (H) = 0S1 (H) + (1 + r)(X0 ; 0S0 ) = 5:59 V1 (H) = 5:60 X1 (T ) = 0S1 (T) + (1 + r)(X0 ; 0S0 ) = 0:01 V1 (T ) = 0 X1 (HH) = 1(H)S1 (HH) + (1 + r)(X1 (H) ; 1(H)S1 (H)) = 11:01 V1 (HH) = 11
etc.
Example 4.2 (European Call) Let u = 2 with expiration time 2 and payoff function
~ ~ d = 1 r = 1 S0 = 4 p = q = 2 4 V2 = (S2 ; 5)+ :
1 2,
Note that
V2 (HH) = 11 V2 (HT) = V2 (TH) = 0 V2 (T T ) = 0 1 V1 (H) = 4 2 :11 + 1 :0] = 4:40 2 5 41 V1(T) = 5 2 :0 + 1 :0] = 0 2 41 V0 = 5 2 4:40 + 1 0] = 1:76: 2 Dene vk (x) to be the value of the call at time k when Sk = x. Then v2 (x) = (x ; 5)+ v1(x) = 4 1 v2 (2x) + 1 v2 (x=2)] 2 52 v0(x) = 4 1 v1 (2x) + 1 v1 (x=2)]: 2 52
69
Let
v2(16) = 11 v2 (4) = 0 v2 (1) = 0 1 v1(8) = 4 2 :11 + 1 :0] = 4:40 2 5 1 v1(2) = 4 1 :0 + 2 :0] = 0 52 1 v0 = 4 2 4:40 + 1 0] = 1:76: 2 5 k (x) be the number of shares in the hedging portfolio at time k when S k = x. Then
k
f V0 = (1 + r) n IEVn
;
and so we could compute V0 by simulation. More specically, we could simulate n coin tosses ! = (!1 : : : !n ) under the risk-neutral probability measure. We could store the value of Vn (! ). We could repeat this several times and take the average value of Vn as an f EV approximation to I n . 2. Approximate a many-period model by a continuous-time model. Then we can use calculus and partial differential equations. Well get to that. 3. Look for Markov structure. Example 4.2 has this. In period 2, the option in Example 4.2 has three possible values v 2 (16) v2(4) v2(1), rather than four possible values V 2(HH ) V2(HT ) V2(TH ) If there were 66 periods, then in period 66 there would be 67 possible stock price values (since the nal price depends only on the number of up-ticks of the stock price i.e., heads so far) and hence only 67 possible option values, rather than 2 66 7 1019.
V2(TT ).
70
F.
Let
(The Markov Property). For each k = 0 1 : : : n ; 1, the distribution of Xk+1 conditioned on F k is the same as the distribution of X k+1 conditioned on X k .
4.3.1
= B(IR), we have
4
< 1, we have
IE euXk+1 F k = IE euXk+1 Xk :
(If we x u and dene h(x) = eux , then the equations in (b) and (c) are the same. However in (b) we have a condition which holds for every function h, and in (c) we assume this condition only for functions h of the form h(x) = e ux . A main result in the theory of Laplace transforms is that if the equation holds for every h of this special form, then it holds for every h, i.e., (c) implies (b).) (d) (Agreement of characteristic functions) For every u 2 IR, we have
where i = 1.)
p;1. (Since jeiuxj = j cos x +sin xj 1 we dont need to assume that IE jeiuxj <
71
Remark 4.1 In every case of the Markov properties where IE : : : jXk ] appears, we could just as well write g (Xk ) for some function g . For example, form (a) of the Markov property can be restated as: For every A 2 B, we have
IR : : : Ak+j IR,
: : : uk+j ) 2 IRj for which IE jeuk+1 Xk+1 +:::+uk+j Xk+j j < 1, we have
h = IA ).
Now take conditional expectation on both sides of the above equation, conditioned on use the tower property on the left, to obtain
(X k ), and
(3.1)
are equal to the RHS of (3.1)), they are equal to each other, and this is property (A) with j
Example 4.3 It is intuitively clear that the stock price process in the binomial model is a Markov process. We will formally prove this later. If we want to estimate the distribution of S k+1 based on the information in F k , the only relevant piece of information is the value of Sk . For example,
(3.2)
is a function of Sk . Note however that form (b) of the Markov property is stronger then (3.2); the Markov property requires that for any function h,
Consider a model with 66 periods and a simple European derivative security whose payoff at time 66 is
73
IP (A \ B) = IP (A)IP (B):
We say that a random variable X is independent of a -algebra G if by X , is independent of G .
(X ), the
-algebra generated
Example 4.4 Consider the two-period binomial model. Recall that F 1 is the -algebra of sets determined by the rst toss, i.e., F 1 contains the four sets
4 4 AH = fHH HT g AT = fTH TT g
fHH T H g fHT T T g
Let H be the -algebra of sets determined by the second toss, i.e., H contains the four sets
: Then F 1 and H are independent. For example, if we take A = fHH HT g from F1 and B = fHH T H g from H, then IP (A \ B) = IP(HH) = p2 and IP (A)IP(B) = (p2 + pq)(p2 + pq) = p2 (p + q)2 = p2:
Note that F 1 and S2 are not independent (unless p = 1 or p = 0). For example, one of the sets in (S2 ) is f! S2 (!) = u2S0 g = fHH g. If we take A = fHH HT g from F 1 and B = fHH g from (S2 ), then IP(A \ B) = IP(HH) = p2 , but
The following lemma will be very useful in showing that a process is Markov: Lemma 4.15 (Independence Lemma) Let X and Y be random variables on a probability space ( F P). Let G be a sub- -algebra of F . Assume
74
X is independent of G ; Y is G -measurable.
Let f (x
Then
IE f (X Y )jG] = g (Y ):
Remark. In this lemma and the following discussion, capital letters denote random variables and lower case letters denote nonrandom variables.
Example 4.5 (Showing the stock price process is Markov) Consider an n-period binomial model. Fix a
4 k 4 = SS+1 and G = F k . Then X = u if !k+1 = H and X = d if !k+1 = T . Since X k 4 depends only on the (k + 1)st toss, X is independent of G . Dene Y = Sk , so that Y is G -measurable. Let h 4 be any function and set f(x y) = h(xy). Then 4 g(y) = IEf(X y) = IEh(Xy) = ph(uy) + qh(dy):
time k and dene X The Independence Lemma asserts that
IE h(Sk+1 )jF k ] = = = =
This shows the stock price is Markov. Indeed, if we condition both sides of the above equation on (S k ) and use the tower property on the left and the fact that the right hand side is (S k )-measurable, we obtain
IE h(Sk+1 )jSk ] = ph(uSk ) + qh(dSk ): Thus IE h(Sk+1 )jF k ] and IE h(Sk+1 )jXk ] are equal and form (b) of the Markov property is proved. IE h(Sk+1 )jF k ] as a function of Sk . This is a special case of Remark 4.1.
Not only have we shown that the stock price process is Markov, but we have also obtained a formula for
Consider a simple European derivative security with payoff at time n of v n (Sn Examples:
Mk = 1maxk Sj : j
4
Mn ) .
75
vn(Sn Mn ) = (Mn ; K )+ (Lookback option); vn(Sn Mn ) = IMn B (Sn ; K )+ (Knock-in Barrier option).
Lemma 5.16 The two-dimensional process f(S k Mk )gn=0 is Markov. (Here we are working under k the risk-neutral measure I , although that does not matter). P Proof: Fix k. We have
Mk+1 = Mk _ Sk+1
4
f IP (Z = u) = p IP (Z = d) = q ~ f ~
and Z is independent of F k . Let h(x
k = SS+1 , so k
1 f Vk = 1 + r IE Vk+1jF k ] k = 0 1 : : : n ; 1:
At the nal time, we have
Vn = vn (Sn Mn):
Vn
Mn 1)] :
;
Vn 1 = vn 1 (Sn
; ;
Mn 1):
;
Chapter 5
vn (x) = g (x) 1 ~ vk (x) = 1 + r pvk+1 (ux) + qvk+1(dx)]: ~ Then vk (Sk ) is the value of the option at time k, and the hedging portfolio is given by vk+1 (uSk ) ; vk+1 (dSk ) k = 0 1 2 : : : n ; 1: k= (u ; d)Sk
Now consider an American option. Again a function g is specied. In any period k, the holder of the derivative security can exercise and receive payment g (Sk ). Thus, the hedging portfolio should create a wealth process which satises
Xk g (Sk) 8k
almost surely.
This is because the value of the derivative security at time k is at least g (S k ), and the wealth process value at that time must equal the value of the derivative security. American algorithm.
vn(x) = g (x) 1 p vk (x) = max 1 + r (~vk+1 (ux) + q vk+1 (dx)) g(x) ~ Then vk (Sk ) is the value of the option at time k.
77
78
S2 (HH) = 16 v2 (16) = 0
v (4) = 1 2
S2 (TT) = 1
v (1) = 4 2
Figure 5.1: Stock price and nal value of an American put option with strike price 5.
Example 5.1 See Fig. 5.1. S0 Then
1 ~ ~ 2 = 4 u = 2 d = 1 r = 4 p = q = 1 n = 2. Set v2 (x) = g(x) = (5 ; x)+ . 2
v1 (8) = max 4 1 :0 + 1 :1 (5 ; 8)+ 2 5 2 = max 2 0 5 = 0:40 v1 (2) = max 4 1 :1 + 1 :4 (5 ; 2)+ 2 5 2 = maxf2 3g = 3:00 v0 (4) = max 4 1 :(0:4) + 1 :(3:0) (5 ; 4)+ 2 5 2 = maxf1:36 1g = 1:36
Let us now construct the hedging portfolio for this option. Begin with initial wealth X 0 0 as follows:
= 1:36.
Compute
0:40 = = = = 3:00 = = = =
v1 (S1 (H)) S1 (H) 0 + (1 + r)(X0 ; 0S0 ) 8 0 + 5 (1:36 ; 4 0) 4 3 0 + 1:70 =) 0 = ;0:43 v1 (S1 (T)) S1 (T) 0 + (1 + r)(X0 ; 0S0 ) 2 0 + 5 (1:36 ; 4 0) 4 ;3 0 + 1:70 =) 0 = ;0:43
79
1 = = = = 4 = = = =
We get different answers for
1
v2 (4) S2 (TH) 1(T) + (1 + r)(X1 (T ) ; 1(T)S1 (T)) 5 4 1(T ) + 4 (3 ; 2 1(T)) 1:5 1(T ) + 3:75 =) 1(T) = ;1:83 v2 (1) S2 (TT ) 1(T ) + (1 + r)(X1 (T) ; 1 (T )S1 (T )) 5 1(T) + (3 ; 2 1(T)) 4 ;1:5 1(T ) + 3:75 =) 1(T ) = ;0:16 (T)! If we had X1 (T ) = 2, the value of the European put, we would have 1 = 1:5 1(T ) + 2:5 =) 4 = ;1:5 1(T) + 2:5 =)
1
(T) = ;1
1
(T) = ;1
Xk+1 =
=
Here, Ck is the amount consumed at time k. The discounted value of the portfolio is a supermartingale. The value satises Xk
g(Sk) k = 0 1 : : : n.
The value process is the smallest process with these properties. When do you consume? If
f IE ((1 + r)
or, equivalently,
80 and the holder of the American option does not exercise, then the seller of the option can consume to close the gap. By doing this, he can ensure that X k = vk (Sk ) for all k, where vk is the value dened by the American algorithm in Section 5.1. In the previous example, v1 (S1(T )) = 3 v2(S2(TH )) = 1 and v2(S2 (TT )) = 4. Therefore,
= = v1 (S1(T )) =
4 5 5 2 2 3
so there is a gap of size 1. If the owner of the option does not exercise it at time one in the state ! 1 = T , then the seller can consume 1 at time 1. Thereafter, he uses the usual hedging portfolio
In the example, we have v1 (S1(T )) = g (S1(T )). It is optimal for the owner of the American option to exercise whenever its value vk (Sk ) agrees with its intrinsic value g (S k ). Denition 5.1 (Stopping Time) Let ( F tion. A stopping time is a random variable
f! 2
P) be a probability space and let fF k gn=0 be a ltrak : !f0 1 2 : : : ng f1g with the property that: (!) = kg 2 F k 8k = 0 1 : : : n 1:
q= ~
1. 2
Example 5.2 Consider the binomial model with n = 2 S 0 = 4 u = 2 d = 1 r = 1 , so p = ~ 2 4 v0 v1 v2 be the value functions dened for the American put with strike price 5. Dene
Let
(!) =
We verify that is indeed a stopping time:
1 2
if ! if !
2 AT 2 AH
Example 5.3 (A random time which is not a stopping time) In the same binomial model as in the previous example, dene
81
80 < (!) = : 1 2
stops when the stock price reaches its minimum value. This if ! if ! if !
2 AH
= TH = TT
A \ f! (!) = kg 2 F k 8k:
= minfk vk (Sk ) = (5 ; Sk )+ g
i.e.,
(!) =
The set fHT g is determined by time , but the set fTH g is not. Indeed,
1 2
if ! if !
2 AT 2 AH
Notation 5.1 (Value of Stochastic Process at a Stopping Time) If ( F P) is a probability space, fF k gn=0 is a ltration under F , fXk gn=0 is a stochastic process adapted to this ltration, and is k k a stopping time with respect to the same ltration, then X is an F -measurable random variable whose value at ! is given by
82 Theorem 3.17 (Optional Sampling) Suppose that fYk F k g1 (or fYk F k gn=0 ) is a submartink=0 k gale. Let and be bounded stopping times, i.e., there is a nonrandom number n such that
n
If almost surely, then
almost surely.
Y IE (Y jF ): Taking expectations, we obtain IEY IEY , and in particular, Y 0 = IEY0 IEY is a supermartingale, then implies Y IE (Y jF ). implies Y = IE (Y jF ). If fYk F k gk=0 is a martingale, then
1
. If fYk
F k gk=0
1
Example 5.5 In the example 5.4 considered earlier, we dene (!) = 2 for all ! 2 . Under the risk-neutral probability measure, the discounted stock price process ( 5 );k Sk is a martingale. We compute 4
e IE
The atoms of F are fHH g
"
4 5
S2 F :
fHT g
e IE e IE
and for !
"
and AT . Therefore,
"
4 2 S F (HH) = 5 2 # 4 2 S F (HT) = 5 2
4 5 4 5
S2 (HH) S2 (HT )
2
2 AT ,
e IE
"
4 5
S2 F (!) =
S2 (T T)
e IE
"
4 5
S2 F (!) = 4 5
( )
S (! ) (!):
83
(4/5) S (H) = 6.40 1 (16/25) S2 (HT) = 2.56 S =4 0 (16/25) S2 (TH) = 2.56 (4/5) S (T) = 1.60 1
84
Chapter 6
f Vk = max (1 + r)k IE (1 + r) G jF k ]
;
satisfying
k almost surely.
(b) The discounted value process f(1 + r);k Vk gn=0 is the smallest supermartingale which satises k
Vk Gk 8k
almost surely.
which satises
f V0 = IE (1 + r) G ]
;
= minfk Vk = Gk g
4
k (!1
86 (e) Suppose for some k and ! , we have Vk (!) = Gk (!). Then the owner of the derivative security should exercise it. If he does not, then the seller of the security can immediately consume
f Vk = (1 + r)k max IE (1 + r) G jF k ] : T
4 ; 2
Vk Gk for every k.
2 Tk to be the constant k.
attain the maximum in the denition of V k+1 , i.e.,
Lemma 2.19 The process f(1 + r);k Vk gn=0 is a supermartingale. k Proof: Let
fh (1 + r) (k+1) Vk+1 = IE (1 + r)
; ;
G jF k+1 :
;
Because
is also in T k , we have
f IE (1 + r)
(k+1) Vk+1jF k ]
f f IE IE (1 + r) G jF k+1]jF k f IE (1 + r) G jF k ] f max IE (1 + r) G jF k ] Tk = (1 + r) k Vk :
= =
; ; 2 ;
Yk Gk k = 0 1 : : : n
and f(1 + r);k Yk gn=0 is a supermartingale, then k
a.s.,
Yk Vk k = 0 1 : : : n
a.s.
87
f IE (1 + r) Y jF k ] (1 + r) k Yk 8 2 Tk :
; ;
Therefore,
f Vk = (1 + r)k max IE (1 + r) G jF k ] T
;
= Yk :
Lemma 2.21 Dene
1 f Ck = Vk ; 1 + r IE Vk+1 jF k ] n f = (1 + r)k (1 + r) k Vk ; IE (1 + r)
;
(k+1) Vk+1jF k ]
Since f(1 + r);k Vk gn=0 is a supermartingale, C k must be non-negative almost surely. Dene k
k (!1
Set X0
Then
Xk = Vk 8k:
for some
We proceed by induction on k. The induction hypothesis is that X k = Vk k 2 f0 1 : : : n ; 1g, i.e., for each xed (!1 : : : !k) we have Xk (!1 : : : !k ) = Vk (!1 : : : !k ): Proof: We need to show that
88 Since (!1 : : : !k ) will be xed for the rest of the proof, we will suppress these symbols. For example, the last equation can be written simply as
1 p Vk ; Ck = 1 + r (~Vk+1 (H ) + qVk+1 (T )) : ~
We compute
Xk+1 (H ) =
= = = =
k Sk+1 (H ) + (1 + r)(Xk ; Ck ; k Sk ) Vk+1 (H ) ; Vk+1(T ) (S (H ) ; (1 + r)S ) k Sk+1 (H ) ; Sk+1 (T ) k+1 +(1 + r)(Vk ; Ck ) Vk+1 (H ) ; Vk+1(T ) (uS ; (1 + r)S ) k k (u ; d)Sk +~Vk+1 (H ) + q Vk+1 (T ) p ~ (Vk+1 (H ) ; Vk+1 (T ))~ + pVk+1(H ) + q Vk+1 (T ) q ~ ~ Vk+1 (H ):
Hedging a short position (one payment). Here is how we can hedge a short position in the j th European derivative security. The value of European derivative security j at time k is given by
f Vk(j) = (1 + r)k IE (1 + r) j Cj jF k ] k = 0 : : : j
;
(j ) (! k 1
Thus, starting with wealth V 0 , and using the portfolio ( 0 time j we have wealth Cj .
:::
Hedging a short position (all payments). Superpose the hedges for the individual payments. In P (j) other words, start with wealth V 0 = n=0 V0 . At each time k 2 f0 1 : : : n ; 1g, rst make the j payment Ck and then use the portfolio
89
corresponding to all future payments. At the nal time n, after making the nal payment Cn , we will have exactly zero wealth. Suppose you own a compound European derivative securityfC j gn=0 . Compute j
V0 =
n X j =0
2n 3 X f V0(j ) = IE 4 (1 + r) j Cj 5
;
j =0
;1 and the hedging portfolio is f k gn=0 . You can borrow V0 and consume it immediately. This leaves k you with wealth X 0 = ;V0 . In each period k, receive the payment Ck and then use the portfolio ; k . At the nal time n, after receiving the last payment Cn, your wealth will reach zero, i.e., you will no longer have a debt.
Cj = I
=j
Gj :
In other words, once he chooses a stopping time, the owner has effectively converted the American derivative security into a compound European derivative security, whose value is
V0( )
2n 3 X f = IE 4 (1 + r) j Cj 5 2j=0 3 n X f = IE 4 (1 + r) j I =j Gj 5
; ;
f = IE (1 + r) G ]:
;
j =0
The owner of the American derivative security can borrow this amount of money immediately, if he chooses, and invest in the market so as to exaclty pay off his debt as the payments fC j gn=0 are j received. Thus, his optimal behavior is to use a stopping time Lemma 4.22 which maximizes V 0 .
( )
90
f Let 0 be a stopping time which maximizes V0 , i.e., V0 = I (1 + r); G 0 : Because f(1 + r);k Vk gn=0 E k is a supermartingale, we have from the optional sampling theorem and the inequality V k Gk , the following:
( )
0
V0
f IE (1 + r) 0 V 0 jF 0 i fh = IE (1 + r) 0 V 0 i fh IE (1 + r) 0 G 0 = V0:
; ; ; ;
Therefore, and
f f V0 = IE (1 + r) 0 V 0 = IE (1 + r) 0 G 0
;
V 0=G
0
0 a.s.
f V0 = max IE (1 + r) G ] T
; 2
(4.1)
V 0=G
0 a.s.
= minfk Vk = Gk g
;
(1 + r)
= (1 + r) V i fh IE (1 + r) 0 V 0 jF i fh = IE (1 + r) 0 G 0 jF :
; ;
f IE (1 + r)
i fh i IE (1 + r) 0 G 0 = V0 :
;
It follows that also attains the maximum in (4.1), and is therefore an optimal exercise time for the American derivative security.
Chapter 7
Jensens Inequality
7.1 Jensens Inequality for Conditional Expectations
Lemma 1.23 If ' : IR!IR is convex and IE j'(X )j < 1, then
=f
g '(x) = x2:
IEX 2 (IEX )2:
Proof: Since ' is convex we can express it as follows (See Fig. 7.1):
IE '(X )jG]
This implies
IE '(X )jG]
= '(IE X jG]):
h is linear
91
92
Figure 7.1: Expressing a convex function as a max over linear functions. Theorem 1.24 If fYk gn=0 is a martingale and k Proof: is convex then f'(Y k )gn=0 is a submartingale. k
IE '(Yk+1)jF k ]
where the LHS is the European value and the RHS is the American value. In particular optimal exercise time. Proof: Because g is convex, for all
= n is an
g (x):
93
( x, g(x))
(x,g(x))
x ( x, g( x))
Figure 7.2: Proof of Cor. 2.25 Therefore,
1 g 1 + r Sk+1
1 g (S ) 1 + r k+1
;
and
1 g (S )jF 1 + r k+1 k 1 f (1 + r) k IE g 1 + r Sk+1 jF k f 1 (1 + r) k g IE 1 + r Sk+1 jF k = (1 + r) k g (Sk ) So f(1 + r) k g (Sk )gn=0 is a submartingale. Let be a stopping time satisfying 0 k
f IE (1 + r)
(k+1) g (Sk+1)jF k
f = (1 + r) k IE
; ;
n. The
f (1 + r) g (S ) IE (1 + r) n g (Sn)jF ] :
; ; ;
f IE (1 + r) g(S )]
f f IE IE (1 + r) n g(Sn)jF ] f = IE (1 + r) n g (Sn)] :
; ; ;
and this last expression is the value of the European derivative security. Of course, the LHS cannot be strictly less than the RHS above, since stopping at time n is always allowed, and we conclude that
94
S2 (HH) = 16
S2 (TT) = 1
fY k gn=0 the k
^
Yk (!) (!) k = 0 1 : : : n:
(!) = 1 2
if if
Example 7.1 (Stopped Process) Figure 7.3 shows our familiar 3-period binomial example. Dene
Then
!1 = T !1 = H:
if if if if
! = HH ! = HT ! = TH ! = T T:
Theorem 3.26 A stopped martingale (or submartingale, or supermartingale) is still a martingale (or submartingale, or supermartingale respectively). Proof: Let fYk gn=0 be a martingale, and be a stopping time. Choose some k 2 f0 1 k The set f kg is in F k , so the set f k + 1g = f kgc is also in F k . We compute
: : : ng.
95
96
Chapter 8
Random Walks
8.1 First Passage Time
Toss a coin innitely many times. Then the sample space is the set of all innite sequences ! = (!1 !2 : : : ) of H and T . Assume the tosses are independent, and on each toss, the probability of H is 1 , as is the probability of T . Dene 2
Yj (!) = Mk =
;1
k X
if if
!j = H !j = T
M0 = 0
The process fMk g1 is a symmetric random walk (see Fig. 8.1) Its analogue in continuous time is k=0 Brownian motion. Dene
j =1
Yj k = 1 2 : : :
= minfk 0 Mk = 1g: If Mk never gets to 1 (e.g., ! = (TTTT : : : )), then = 1. The random variable
8.2 is almost surely nite
is called the rst passage time to 1. It is the rst time the number of heads exceeds by one the number of tails.
Nk = exp Mk ; k log e + e 2
= e Mk 2 e +e
97
!)
98
Mk k
e e + 2 1
Figure 8.2: Illustrating two functions of
2 e e + 1
are martingales. (Take Mk = ;Sk in part (i) of the Homework Problem and take (v).) Since N0 = 1 and a stopped martingale is a martingale, we have
=;
in part
1 = IENk = IE e Mk
^
"
2 e +e
(2.1)
for every xed 2 IR (See Fig. 8.2 for an illustration of the various functions involved). We want to let k!1 in (2.1), but we have to worry a bit that for some sequences ! 2 , (! ) = 1. We consider xed As k!1,
> 0, so
e +e
2 e +e
k
; ^
< 1:
if if
(
0
2 e +e;
<1
Furthermore, Mk^
=1
99
0 e Mk lim e Mk k!
1
2 e +e
k
^
e:
if if
In addition,
^
2 e +e
2 e +e;
<1 = 1:
IE e Mk^
"
2 e +e
=1 = 1:
2 IE e e + e
For all
<
1g
(2.2)
2 (0 1], we have
so we can let
2 0 e e +e
<
1g
i.e.,
We know there are paths of the symmetric random walk fMk g1 which never reach level 1. We k=0 have just shown that these paths collectively have no probability. (In our innite sample space , each path individually has zero probability). We therefore do not need the indicator I f < 1g in (2.2), and we rewrite that equation as
IE
e +e
=e :
;
(2.3)
Solution:
e + e ;2=0 (e )2 ; 2e + = 0
; ; ;
100
e =1
;
1;
:
2
We want
e = 1; 1;
;
: > 0:
2
IE
With
e +e
;
=e
= 1; 1; 2 = e +e
;
this becomes
IE
= 1; 1;
0 < < 1:
(3.1)
We have computed the moment generating function for the rst passage time to 1.
8.4 Expectation of
Recall that
IE
so
= 1; 1; = IE (
0< <1
;
d IE d
= 1 ;p 1 ; 2 1;
= dd 1 ; 1 ;
1)
2 2 :
101
IE (
to obtain
1) =
1 ;p 1 ; 2 1;
IE = 1:
= minfk Mk = 1g
4
Thus in summary:
IP f < 1g = 1 IE = 1:
8.5 The Strong Markov Property
The random walk process fMk g1 is a Markov process, i.e., k=0
IE
= IE
In discrete time, this Markov property implies the Strong Markov property:
IE
= IE
M +1 M +2 : : : j F ] j M ]:
Then 2 ; 1 is the number of periods between the rst arrival at level 1 and the rst arrival at level 2. The distribution of 2 ; 1 is the same as the distribution of 1 (see Fig. 8.3), i.e.,
m = minfk
4
0 Mk = mg m = 1 2 : : :
IE
2; 1
= 1; 1;
2 (0 1):
102
Mk k 1 2 1
Figure 8.3: General rst passage times. For
2 (0 1),
IE
jF 1 ] = IE
=
1
2; 1 2; 1
IE IE
jF 1 jF 1 ]
= = =
Take expectations of both sides to get
jM 1 ]
IE 2 1 ] (M 1 = 1 not random ) ! p
1
;
1; 1;
IE
= IE 1 : 1 ; 1 ; = 1; 1;
! 2 2
In general,
IE
= 1; 1;
! 2 m
2 (0 1):
r=
Sk = S0 uMk
103 Suppose
where
= minfk Mk = ;1g:
4
= minfk Mk = ;2g:
4
f f Because the random walk is symmetric under I , ;m has the same distribution under I as the P P stopping time m in the previous section. This observation leads to the following computations of value. Value of Rule 1: f V ( ;1) = IE (1 + r) ;1 (5 ; S ;1 )+ h4 i = (5 ; 2)+IE ( 5 ) ;1
;
= 3: 3 = 2:
Value of Rule 2:
1 ; 1 ; ( 4 )2 5
4 5
f 5 V ( ;2 ) = (5 ; 1)+IE ( 4 ) ;2 1 = 4:( 2 )2 = 1:
Suppose instead we start with S 0 = 8, and stop the rst time the price falls to 2. This requires 2 down steps, so the value of this rule with this initial stock price is
This suggests that the optimal rule is Rule 1, i.e., stop (exercise the put) as soon as the stock price falls to 2, and the value of the put is 3 if S0 = 4. 2
In general, if S0 = 2j for some j 1, and we stop when the stock price falls to 2, then j ; 1 down steps will be required and the value of the option is
fh 4 i (5 ; 2)+ IE ( 5 ) ;2 = 3:( 1 )2 = 3 : 2 4 h i
;
v(2j ) = 3:( 1 )j 2
4
j = 1 2 3 :::
104
1, then the initial price is at or below 2. In this case, we exercise If S0 = 2j for some j immediately, and the value of the put is
v (2j ) = 5 ; 2j j = 1 0 ;1 ;2 : : :
4
Proposed exercise rule: Exercise the put whenever the stock price is at or below 2. The value of this rule is given by v (2 j ) as we just dened it. Since the put is perpetual, the initial time is no different from any other time. This leads us to make the following: Conjecture 1 The value of the perpetual put at time k is v (S k ). How do we recognize the value of an American derivative security when we see it? There are three parts to the proof of the conjecture. We must show: (a)
v (Sk ) (5 ; Sk )+ 8k o n (b) ( 4 )k v (Sk ) is a supermartingale, 5 k=0 (c) fv (Sk )gk=0 is the smallest process with properties (a) and (b).
1 1
Note: To simplify matters, we shall only consider initial stock prices of the form S 0 always of the form 2j , with a possibly different j . Proof: (a). Just check that
= 2j , so Sk is
v(2j ) = 3:( 1 )j 2
4 4
(5 ; 2j )+
for for
j 1 j 1:
v(2j ) = 5 ; 2j (5 ; 2j )+
This is straightforward. Proof: (b). We must show that
v(Sk )
By assumption, Sk
= 2j for some j . We must show that 2 v(2j ) 5 v (2j +1) + 2 v (2j 1): 5 If j 2, then v (2j ) = 3:( 1 )j 1 and 2 2 v (2j +1 ) + 2 v (2j 1) 5 5 1 1 = 2 :3:( 2 )j + 2 :3:( 2 )j 2 5 5 2 = 3: 5 : 1 + 2 ( 1 )j 2 4 5 2 = 3: 1 :( 1 )j 2 2 2 = v (2j ):
; ; ; ; ; ;
f 5 IE 4 v (Sk+1)jF k = 4 : 1 v (2Sk ) + 4 : 1 v ( 1 Sk ): 5 2 5 2 2
105
If j
There is a gap of size 1. This concludes the proof of (b). Proof: (c). Suppose fYk gn=0 is some other process satisfying: k (a) (b)
Yk (5 ; Sk )+ 8k f( 4 )k Yk gk=0 is a supermartingale. 5
1
(7.1)
Actually, since the put is perpetual, every time k is like every other time, so it will sufce to show (7.2)
provided we let S0 in (7.2) be any number of the form 2j . With appropriate (but messy) conditioning on F k , the proof we give of (7.2) can be modied to prove (7.1).
1,
so if S0
v (2j ) = 5 ; 2j = (5 ; 2j )+ Y0 (5 ; 2j )+ = v (S0):
106 Then
IE ( 4 ) (5 ; S )+ = v(S0): 5
Comment on the proof of (c): If the candidate value process is the actual value of a particular exercise rule, then (c) will be automatically satised. In this case, we constructed v so that v (S k ) is the value of the put at time k if the stock price at time k is S k and if we exercise the put the rst time (k, or later) that the stock price is 2 or less. In such a situation, we need only verify properties (a) and (b).
v (x) (K ; x)+ 8x, 1 ~ ~ (b) v (x) 1+r pv (ux) + q v (dx)] 8x (c) At each x, either (a) or (b) holds with equality.
(a) In the example we worked out, we have For
v (x) =
We then have (see Fig. 8.4): (a) (b)
x 3 5 ; x 0 < x 3:
6 x
107
v(x) 5
(3,2)
If 3 < x < 4 or 4 < x < 6, then both (a) and (b) are strict. This is an artifact of the discreteness of the binomial model. This artifact will disappear in the continuous model, in which an analogue of (a) or (b) holds with equality at every point.
= 0. Dening
= minfk 0 Mk = 1g
we recall that
IE
= 1; 1;
0 < < 1:
We will use this moment generating function to obtain the distribution of . We rst obtain the Taylor series expasion of IE as follows:
108
f (x) = 1 ; 1 ; x f (0) = 0 1 1 1 f (x) = 2 (1 ; x) 2 f (0) = 2 3 f (x) = 1 (1 ; x) 2 f (0) = 1 4 4 3 (1 ; x) 5 f (0) = 3 2 f (x) = 8 8 ::: : f (j)(x) = 1 3 : :2j (2j ; 3) (1 ; x) f (j)(0) = 1 3 : : : (2j ; 3)
0 ; 0 00 ; 00 000 ; 000
(2j 1) 2
2j
j =0
X
1
X = x+ 2
1
j =1
1 2j 1 2
;
j =2
1 2j 1 1 2 (j ; 1)
2j ; 2 xj : j
So we have
IE
= 1; 1; = 1 f ( 2) = 2 +
X
1
2j 1
;
j =2
1
2
2j 1 IP f
;
(j ; 1)
2j ; 2 :
But also,
IE =
X
j =1
= 2j ; 1g:
109
= 2.
IP f = 1g = IP f = 2j ; 1g =
1 2
! 1 2j 1 1 2j ; 2 j = 2 3 : : : 2 (j ; 1) j
;
= 1 and
IP f
For j
;3g
2,
IP f = 2j ; 1g = IP f
= = =
= = = =
2j ; 1g ; IP f 2j ; 3g 1 ; IP fM2j 1 = ;1g] ; 1 ; IP fM2j 3 = ;1g] IP fM2j 3 = ;1g ; IP fM2j 1 = ;1g (2j ; 3)! ; 1 2j 1 (2j ; 1)! 1 2j 3 2 (j ; 1)!(j ; 2)! 2 j !(j ; 1)! 2j 1 (2j ; 3)! 1 2 j !(j ; 1)! 4j (j ; 1) ; (2j ; 1)(2j ; 2)] 1 2j 1 (2j ; 3)! 2j (2j ; 2) ; (2j ; 1)(2j ; 2)] 2 j !(j ; 1)! 2j 1 (2j ; 2)! 1 2 j !(j ; 1)! ! 2j ; 2 : 1 2j 1 1 2 (j ; 1) j
; ; ; ; ; ; ; ; ; ;
Chapter 9
f P Theorem 1.27 (Radon-Nikodym) Let I and I be two probability measures on a space ( F ). P f Assume that for every A 2 F satisfying IP (A) = 0, we also have I (A) = 0. Then we say that P f is absolutely continuous with respect to I . Under this assumption, there is a nonegative random P IP variable Z such that f IP (A) = Z f P P and Z is called the Radon-Nikodym derivative of I with respect to I . f IEX = IE XZ ] for every random variable X for which IE jXZ j < 1. f Remark 9.2 If I is absolutely continuous with respect to I , and I is absolutely continuous with P P P f, we say that I and IP are equivalent. I and IP are equivalent if and only if f f respect to I P P P f IP (A) = 0 exactly when IP (A) = 0 8A 2 F : 1 f f If I and I are equivalent and Z is the Radon-Nikodym derivative of I w.r.t. I , then Z is the P P P P f Radon-Nikodym derivative of I w.r.t. I , i.e., P P f (1.2) IEX = IE XZ ] 8X f 1 IEY = IE Y: Z ] 8Y: (Let X and Y be related by the equation Y = XZ to see that (1.2) and (1.3) are the same.)
111 (1.3) Remark 9.1 Equation (1.1) implies the apparently stronger condition
ZdIP 8A 2 F
(1.1)
112
Example 9.1 (Radon-Nikodym Theorem) Let = fHH HT TH T T g, the set of coin toss sequences of length 2. Let P correspond to probability 1 for H and 2 for T , and let IP correspond to probability 1 for 3 3 2
f f P P so that I and I are equivalent. The Radon-Nikodym derivative of I with respect to I is P P f IP (!) Z (!) = IP (! ) :
Dene the I -martingale P
Zk = IE Z jF k ] k = 0 1 : : : n:
4
Note that Lemma 2.28 implies that if X is F k -measurable, then for any A 2 F k ,
or equivalently,
f XdIP =
Z
A
XZk dIP:
113
Figure 9.1: Showing the Zk values in the 2-period binomial model example. The probabilities shown f are for I , not I . P P Lemma 2.29 If X is F k -measurable and 0
j k, then
j
1 f IE X jF j ] = Z IE XZk jF j ]:
ZA
A
XZk dIP
f XdIP
Example 9.2 (Radon-Nikodym Theorem, continued) We show in Fig. 9.1 the values of the martingale Zk . We always have Z0 = 1, since
Z0 = IEZ =
e ZdIP = IP ( ) = 1:
= (1 + r) k Zk k = 0 : : : n:
;
114 We then have the following pricing formulas: For a Simple European derivative security with payoff Ck at time k,
f V0 = IE (1 + r) k Ck h i = IE (1 + r) k Zk Ck = IE k Ck ]:
; ;
(Lemma 2.28)
j k,
f Vj = (1 + r)j IE (1 + r) k Ck jF j j h i = (1 + r) IE (1 + r) k Zk Ck jF j Zj 1 IE C jF ] = k k j
; ;
(Lemma 2.29)
Remark 9.3
T0
f Vj = (1 + r)j sup IE (1 + r) G jF j ]
= 1 sup IE G jF j ]:
j
2
j n,
1 = (1 + r)j sup Z IE (1 + r) Z G jF j ]
2
Tj
Tj j
Tj
115
() = 0.36 2 S2 (TT) = 1
f Figure 9.2: Showing the state price values k . The probabilities shown are for I , not I . P P
(c)
We interpret k by observing that k (! )IP (! ) is the value at time zero of a contract which pays $1 at time k if ! occurs.
Example 9.3 (Radon-NikodymTheorem, continued) We illustrate the use of the valuation formulas for 1 European and American derivative securities in terms of market probabilities. Recall that p = 3 , q = 2 . The 3 state price values k are shown in Fig. 9.2. For a European Call with strike price 5, expiration time 2, we have
V2 (HH) = 11 2(HH)V2 (HH) = 1:44 11 = 15:84: V2 (HT) = V2(TH) = V2 (TT ) = 0: V0 = 1 1 15:84 = 1:76: 3 3 2 (HH) V2(HH) = 1:44 11 = 1:20 11 = 13:20 1:20 1 (HH) V1 (H) = 1 13:20 = 4:40 3
Compare with the risk-neutral pricing formulas:
116
+ (5 - S (HH)) = 0 2 + (HH) (5 - S (HH)) = 0 2 2
1/3
+ (5-S 0) =1 + 0 (5-S 0) =1
2/3 1/3
2/3
Figure 9.3: Showing the values k (5 ; Sk )+ for an American put. The probabilities shown are for f I , not I . P P
(2) If we stop at time 2, the value is
1 3
2 3
2 0:72 + 3 1 0:72 + 2 3 3
2 3
1:44 = 0:96
k pk = 1 + r;; dk ~ u d k k
; qk = uk u (1 + rk) : ~ ;d
and for 2
equivalent. Dene
f IP !k+1 = H jF k ] = pk ~ f IP !k+1 = T jF k ] = qk : ~ Let I be the market probability measure, and assume IP f! g > 0 8! 2 P f IP (!) Z (!) = IP (! ) 8! 2
k n,
f IP f!1 = H g = p0 ~ f IP f!1 = T g = q0 ~
. Then I and P
f IP are
117
Zk = IE Z jF k ] k = 0 1 : : : n:
We dene the money market price process as follows:
M0 = 1
Note that Mk is Fk;1 -measurable.
Mk = (1 + rk 1)Mk
;
k = 1 : : : n:
1 = M Zk k = 0 : : : n:
; As before the portfolio process is f consists of X 0, the non-random initial wealth, and
Xk+1 =
k Sk+1 + (1 + rk )(Xk ; k Sk )
k = 0 : : : n ; 1:
1 X n M k
k
k=0
f k Sk gn=0 k
We thus have the following pricing formulas:
and
f k Xk gn=0: k
f C Vj = Mj IE Mk F j k 1 IE C jF ] = k k j
j
American derivative security
Tj
118
IP (A) = Q(A) 8A 2 G
f IP (A) = Z
XdQ 8A 2 G :
Y dIP =
Y dQ
if Y = A for some A 2 G , this is just the denition of IP , and the rest follows from the standard f f machine. If A 2 G and IP (A) = 0, then Q(A) = 0, so I (A) = 0. In other words, the measure I P P f. The Radon-Nikodym theorem implies that P is absolutely continuous with respect to the measure I there exists a G -measurable random variable Z such that
f IP (A) =
4
Z dIP 8A 2 G
A
i.e.,
Z
A
X dQ =
Z dIP 8A 2 G :
This shows that Z has the partial averaging property, and since Z is G -measurable, it is the conditional expectation (under the probability measure Q) of X given G . The existence of conditional expectations is a consequence of the Radon-Nikodym theorem.
Chapter 10
IE log Xn :
Remark 10.1 Regardless of the portfolio used by the agent, f k Xk g1 is a martingale under I , so P k=0
IE nXn = X0
Here, (BC) stands for Budget Constraint. Remark 10.2 If is any random variable satisfying (BC), i.e.,
(BC )
IE
= X0
then there is a portfolio which starts with initial wealth X 0 and produces Xn = at time n. To see this, just regard as a simple European derivative security paying off at time n. Then X 0 is its value at time 0, and starting from this value, there is a hedging portfolio which produces X n = . Remarks 10.1 and 10.2 show that the optimal obtained by solving the following Constrained Optimization Problem: Find a random variable which solves: Maximize Subject to Equivalently, we wish to Maximize
IE log
n
IE
= X0:
X
!
2
(log (! )) IP (! )
119
120 Subject to
X
!
2
n (! )
(!)IP (! ) ; X0 = 0:
2n X
k=1
(log xk )IP (! k )
2n X
k=1
n (! k )xk IP (! k )
; Xo = 0:
Theorem 1.30 (Lagrange Multiplier) If (x1 Maxmize Subject to then there is a number such that
g(x1 : : : xm) = 0:
For our problem, (1.1) and (1.2) become
(1.2)
xk IP (! k ) =
2n X
n (! k )IP (! k ) n (! k )xk IP (! k )
k = 1 : : : 2n
= X0 : (1:2 )
0
(1:1 )
0
k=1
xk =
Plugging this into (1.2) we get
n
1 : n (! k )
2 1 X IP (! ) = X =) 1 = X : 0 k 0
k=1
121
0 xk = X! ) k = 1 : : : 2n: ( k n
solves the problem Maximize Subject to
IE log IE ( n ) = X0
= X0 :
n
(1.3)
then (1.4)
1 = Z , i.e.,
(1.5)
IE ( n ) = X0
and let
= X0 :
n
log ;
n X0
log X0 ; 1:
n
1 IE log ; X IE ( n ) IE log ; 1 0
and so
IE log
IE log :
122 In summary, capital asset pricing works as follows: Consider an agent who has initial wealth and wants to invest in the stock and money market so as to maximize
X0
IE log Xn :
The optimal Xn is Xn
0 = Xn , i.e.,
n Xn = X0:
k Xk
so
= IE nXn jF k ] = X0 k = 0 : : : n
Xk = X0
k k (!1
: : : !k ) =
!k T ) : !k T )
Chapter 11
F is a
I is a probability measure on F , i.e., IP (A) is dened for every A 2 F . P A function X : !IR is a random variable if and only if for every Borel subsets of I ), the set R
4 ; 4
-algebra of subsets of .
B 2 B(IR) (the
-algebra of
fX 2 Bg = X 1(B) = f! X (!) 2 Bg 2 F
i.e., X 11.1)
X 1(B) 8B 2 B(IR) where the probabiliy on the right is dened since X 1(B ) 2 F . X is often called the Law of X
X (B ) = IP
; ;
X (B ) =
fX (x) dx 8B 2 B (IR):
123
124
R B
{X B}
where the integral is with respect to the Lebesgue measure on I . fX is the Radon-Nikodym derivaR tive of X with respect to the Lebesgue measure. Thus X has a density if and only if X is absolutely continuous with respect to Lebesgue measure, which means that whenever B 2 B(IR) has Lebesgue measure zero, then
d X (x) = fX (x)dx
IP fX 2 Bg = 0:
11.3 Expectation
Theorem 3.32 (Expectation of a function of X ) Let h : IR!IR be given. Then
IEh(X ) =
4
Z Z
h(X (!)) dIP (!) h(x) d X (x) h(x)fX (x) dx: IR, then these equations are
= =
ZIR
IR
IE 1B (X ) = P fX 2 B g = Z X (B ) = fX (x) dx
B
which are true by denition. Now use the standard machine to get the equations for general h.
125
(X,Y)
C y { (X,Y) C} x
Figure 11.2: Two real-valued random variables X
Y.
XY
induce a
X Y (C ) = IP f(X
4
Y ) 2 C g 8C 2 B(IR2 ):
Y ) is a function fX Y : IR2! 0 1)
that satises
X Y (C ) =
ZZ
C
fX Y
is the Radon-Nikodym derivative of X Y with respect to the Lebesgue measure (area) on IR2 .
IEk(X Y ) =
4
= =
ZZ
IR ZZ
2
y)
IR
126
ZB Z
B
4
fY (y ) dy fX Y (x y ) dx:
IR
where
fY (y ) =
IR
(X Y ) has joint density f X Y . Let h : IR!IR be given. Recall that IE h(X )jY ] = IE h(X )j (Y )] depends on ! through Y , i.e., there is a function g (y ) (g depending on h) such that
4
fY 2 Bg for some
(6.1)
g (Y ) dIP =
h(X ) dIP 8A 2 (Y )
Z Z
IR
1B (Y )g(Y ) dIP =
ZZ
IR2
(6.2)
1B (y)g(y) Y (dy) =
g(y )fY (y ) dy =
y ) 8B 2 B(IR)
(6.3)
Z
B
Z Z
B IR
(6.4)
127
Z
IR
(7.1)
IE h(X )jY ] = g (Y )
(7.2)
Z Z
B | IR
h(x)fX Y (xjy) dx fY (y ) dy =
g (y )
Z Z
2 B(IR)
{z
j
B IR
h(x)fX Y (x y) dxdy:
IE h(X )jY ] = g (Y ):
The function g is often written as
IE h(X )jY = y ] = g (y ) =
Z
IR
In conclusion, to determine IE
IR
h(x)fX Y (xjy ) dx
j
>0
1 2
Let
fX Y (x y) =
1 2
1 p1 ; 2 exp
1 ; 2(1 ; 2 )
x2 ; 2 x y + y2 2 2
1 2
128
The exponent is
fY (y) =
Z1
1 1 Z 1 e; u2 du:e; 2 y = 2 2
2
1 2
1;
;1
; 2(1;1 2)
2 2 2
2 1
x;
1 2
dx:e
1 ; 2 y2 2
;1
Thus Y is normal with mean 0 and variance 2 . 2 Conditional density. From the expressions
; = p1 e 2 2
;2
x;
1 2
dx y , du = p1; 2
fX Y (x y) =
1 2
1 p
1;
e 2
1 ; 2(1; 2 )
1 x 2 1 2
;;
1y 2 2
e; 2
1 y2
2 2
1y 2 fY (y) = p 1 e; 2 2 2 2
we have
fX jY (xjy) = fXfY (x y) Y (y) 2 1 1 p 1 e; 2(1; 2) 12 x ; 21 y : 1 = p 2 1 1; 2 In the x-variable, fX jY (xjy) is a normal density with mean 21 y and variance (1 ; 2 ) Z1 IE X jY = y] = xf (xjy) dx = 1 y 2 ;1 X jY IE
2. 1
Therefore,
"
X;
1 2
Z1
Y =y
1 2
;1 = (1 ; 2 ) 2 : 1
x;
y fX jY (xjy) dx 2
129
IE X jY ] = IE
1 2
(7.3)
"
X;
1 2
Y = (1 ; 2 ) 2: 1
(7.4)
IEX = IE
1 2
IEY = 0
(7.5)
"
X;
1 2
= (1 ; 2 ) 2: 1
(7.6)
Based on Y , the best estimator of X is 21 Y . This estimator is unbiased (has expected error zero) and the expected square error is (1 ; 2 ) 2. No other estimator based on Y can have a smaller expected square error 1 (Homework problem 2.1).
Here, and A is an n
:::
;
A 1 = IE (X ; ):(X ; )T
i.e. the (i j )th element of A;1 is IE (Xi ; i )(Xj ; j ). The random variables in if and only if A;1 is diagonal, i.e.,
X are independent
A 1 = diag(
;
2 2 1 2
:::
2) n
where 2 j
130
= IE (X1 ; 1 )(X2 ; 2 ) :
1 2
;
Thus,
2 A=4
A =
"
;
2 1 1 2
1 2 2 2
#
;
; p
1 (1 2 ) 2) 1 2 (1
2 1
;
;
1 p
1 2 2 2
(1 ) 1 (1 2 )
2
;
3 5
det A =
1 2
1;
and we have the formula from Example 11.1, adjusted to account for the possibly non-zero expectations:
1 p 2 1;
T IEeu :X =
:::
;1
If any n random variables X1 X2 : : : Xn have this moment generating function, then they are jointly normal, and we can read out the means and covariances. The random variables are jointly normal and independent if and only if for any real column vector = (u 1 : : : un )T
Chapter 12
Semi-Continuous Models
12.1 Discrete-time Brownian Motion
Let fYj gn=1 be a collection of independent, standard normal random variables dened on ( j where I is the market measure. As before we denote the column vector (Y1 : : : Yn )T by P therefore have for any real colum vector = (u1 : : : un )T ,
F P), Y. We
If we know Y1 Y2 : : : Yk , then we know B1 B2 : : : Bk . Conversely, if we know B1 then we know Y1 = B1 Y2 = B2 ; B1 : : : Yk = Bk ; Bk;1 . Dene the ltration
j =1
Yj k = 1 : : : n: B2 : : : Bk ,
F0 = f g F k = (Y1 Y2 : : : Yk ) = (B1 B2 : : : Bk ) k = 1 : : : n:
131
132
B k Y2 Y1 0 1 k 2 Y 3 3 Y 4 4
;1
h(y + b)e
1 y2 2
dy:
Sk = S0 exp Bk + ( ; 1 2 )k 2
Note that
k = 0 : : : n:
133
;
1 2
Thus
and
Mk = erk k = 0 1 : : : n:
0
Each
k is F k -measurable.
:::
n 1
;
Wealth process:
Xk+1 Mk+1 =
12.4 Risk-Neutral Measure
Sk+1 Sk Xk Mk+1 ; Mk + Mk :
134
f Theorem 4.36 If I is a risk-neutral measure, then every discounted wealth process P f a martingale under I , regardless of the portfolio process used to generate it. P
Proof:
n Xk on
Mk
k=0
is
f X IE Mk+1 F k k+1
f = IE
=
X = Mk : k
= Vn .
fX f V X0 = IE Mn = IE Mnn : n
Remark 12.1 Hedging in this semi-continuous model is usually not possible because there are not enough trading dates. This difculty will disappear when we go to the fully continuous model.
12.6 Arbitrage
Denition 12.2 An arbitrage is a portfolio which starts with X 0
135
f measure, let X0 = 0, and let Xn be the nal wealth corresponding P Proof: Let I be a risk-neutraln on f is a martingale under I , P to any portfolio process. Since Xk
Mk
k=0
fX fX IE Mn = IE M00 = 0: n
Suppose IP (Xn
(6.1)
0) = 1. We have
(6.2)
f f IP (Xn 0) = 1 =) IP (Xn < 0) = 0 =) IP (Xn < 0) = 0 =) IP (Xn 0) = 1: f (6.1) and (6.2) imply I (Xn P
This is not an arbitrage.
= 0) = 1. We have
If
S S IE Mkk+1 F k = Mkk :IE exp f Yk+1 gjF k ] : expf ; r ; 1 2g 2 +1 S = Mkk : expf 1 2 g: expf ; r ; 1 2g 2 2 Sk : = e r : Mk = r, the market measure is risk neutral. If 6= r, we must seek further.
;
136
Sk+1 = Sk : exp n Yk+1 + ( ; r ; 1 2 )o 2 Mk+1 Mk o n S = Mkk : exp (Yk+1 + r ) ; 1 2 2 n ~ o S = Mkk : exp Yk+1 ; 1 2 2
;
f ~ P We want a probability measure I under which Y1 dom variables. Then we would have
~ Yk+1 = Yk+1 +
r:
i S fh 1 f S ~ IE Mkk+1 F k = Mkk :IE expf Yk+1 gjF k : expf; 2 2g +1 S 1 = Mkk : expf 2 2 g: expf; 1 2g 2 S = Mkk : 3 2n X 1 Z = exp 4 (; Yj ; 2 2 )5 :
j =1
Properties of Z :
Z 0.
: exp ; n 2
= 1:
Dene
f IP (A) =
0 for all A 2 F and
Z
A
Z dIP 8A 2 F :
f Then I (A) P
f IP ( ) = IEZ = 1:
137
2n 3 2n 3 X 5 X IE exp 4 uj Yj = exp 4 1 u2 5 : 2 j
j =1 j =1
i h Z = exp Pn=1 (; Yj ; 1 2) j 2
f IP (A) =
Z
A
Z dIP 8A 2 F
f IEX = IE (XZ ) for every random variable X . f ~ ~ Compute the moment generating function of (Y1 : : : Yn ) under I : P 2n 3 2n 3 n X ~5 X X f IE exp 4 uj Yj = IE exp 4 uj (Yj + ) + (; Yj ; 1 2)5 2 j =1 j =1 j =1 3 2n 3 2n X X = IE exp 4 (uj ; )Yj 5 : exp 4 (uj ; 1 2 )5 2 j =1 j =1 2n 3 2n 3 X1 X 1 = exp 4 2 (uj ; )25 : exp 4 (uj ; 2 2 )5 j =1 2j=1 3 n X 1 2 = exp 4 ( 2 uj ; uj + 1 2 ) + (uj ; 1 2) 5 2 2 j =1 2n 3 X = exp 4 1 u2 5 : 2 j
j =1
138
Sn = S0 exp Bn + ( ; 1 2)n 2
o n 8 n 9 < X = = S0 exp : Yj + ( ; 1 2 )n 2 j =1 8 n 9 < X = = S0 exp : (Yj + r ) ; ( ; r)n + ( ; 1 2)n 2 j =1 8 n 9 < X = 1 ~ = S0 exp : Yj + (r ; 2 2)n : j =1
;
8 n
;1
j =1
rn
b2
Chapter 13
Brownian Motion
13.1 Symmetric Random Walk
Toss a fair coin innitely many times. Dene
Xj (!) = 1
Set
;1
if if
!j = H ! j = T:
M0 = 0 Mk =
k X j =1
Xj
k 1:
1 M !0 k k
k!1:
140 Proof:
(Def. of
Mk :)
(Independence of the Xj s)
1 u 2e k
+ 1e 2
u
u k k
u k
k!
log 1 eux + 1 e ux 2 2
;
u eux ; = xlim0 2 ux ! 1e + 2
= 0:
u e ux 2 1 e ux 2
; ;
Therefore,
k!
1 p Mk ! k
Proof:
Standard normal, as
k!1:
141
u pk
x2 u eux ; u e ux = xlim0 2 1 ux 2 1 ux ! 2x 2 e + 2 e
; ;
1 log 2 eux + 1 e ux 2
;
= xlim0 2 ! = xlim0 ! 1 u2 : =2
Therefore,
u eux ; u e ux 2
;
u2 eux ; u2 e ux 2 2
;
2x
142
k/n
(k+1)/n
Properties of
B(100)(1) :
1 B (100)(1) = 10
100 X
1 IEB(100)(1) = 10
j =1 100 X
Xj IEXj = 0:
var(Xj ) = 1
(Approximately normal)
j =1 100 1 X j =1
B(100)(2) :
1 B (100)(2) = 10
200 X
j =1
Xj
(Approximately normal)
B (100)(1) and B (100)(2) ; B (100)(1) are independent. B (100)(t) is a continuous function of t. To get Brownian motion, let n!1 in B (n) (t) t 0.
13.5 Brownian Motion
(Please refer to Oksendal, Chapter 2.)
143
B (0) = 0, B (t) is a continuous function of t; B has independent, normally distributed increments: If 0 = t0 < t1 < t2 < : : : < tn
and
then
0 s t be given. Then B (s) and B (t) ; B (s) are independent, so B (s) and B (t) = (B (t) ; B (s)) + B (s) are jointly normal. Moreover, IEB (s) = 0 var(B (s)) = s IEB(t) = 0 var(B (t)) = t IEB(s)B (t) = IEB(s) (B(t) ; B(s)) + B(s)] = IEB (s)(B{z) ; B (s)) + IEB 2 (s) (t } | {z } | = s:
0
IEB(s)B(t) = s ^ t:
13.7 Finite-Dimensional Distributions of Brownian Motion
Let be given. Then
2 3 IEB 2(t1 ) IEB (t1 )B (t2) : : : IEB(t1 )B (tn ) 6 7 2 C = 6:IEB:(t:2:):B:(:t1 ) : : : : :IEB: :(t2:): : : : : :: :: :: : : IEB:(:t:2:):B (t:n: )7 6 ::: : :: 4 ::: : ::: : : :7 5 IEB(tn)B (t1 ) IEB(tn )B(t2 ) : : : IEB2 (tn) 2 3 t1 t1 : : : t1 6 7 = 6t1 : : t:2: : ::::::: : : t:2:7 6: : 7 4 5
t1 t2 : : : tn
fF (t)gt
Required properties: For each t, B (t) is F (t)-measurable, For each t and for t < t1
< t2 < < tn , the Brownian motion increments B (t1 ) ; B(t) B(t2 ) ; B(t1 ) : : : B (tn ) ; B (tn 1 ) are independent of F (t).
;
fB(s) 2 C g = f! : B(s !) 2 C g in F (t). Do this for all possible numbers s 2 0 t] and C 2 B(IR).
required by the -algebra properties.
This F (t) contains exactly the information learned by observing the Brownian motion upto time t. fF (t)gt 0 is called the ltration generated by the Brownian motion.
145
s t be given. Then IE B (t)jF (s)] = IE (B(t) ; B(s)) + B(s)jF (s)] = IE B (t) ; B (s)] + B (s) = B (s):
2 IR be given. Then
s t be given. Then
1 = IE Z (s) expf; (B (t) ; B (s)) ; 2 2 (t ; s)g F (s)
146 Let ]k (H ) denote the number of H in the rst k tosses, and let ] k (T ) denote the number of T in the rst k tosses. Then
]k (H ) + ]k (T ) = k ]k (H ) ; ]k (T ) = Mk
which implies,
]k (H ) = 1 (k + Mk ) 2 ]k (T ) = 1 (k ; Mk ): 2
In the nth model, take n steps per unit time. Set S0
S (n)(t) = 1 + pn
1 ; pn
Theorem 10.42 As n!1, the distribution of S (n) (t) converges to the distribution of where B is a Brownian motion. Note that the correction martingale. Proof: Recall that from the Taylor series we have
expf B (t) ; 1 2 tg 2
log(1 + x) = x ; 1 x2 + O(x3) 2
so
+ Mnt
1 2 log(1 + 1 2
pn ) + 1 log(1 ; pn ) 2
pn ) ; 1 log(1 ; pn ) 2
| {z } | {z } !0 !Bt
B(t) ; 1 2t. 2
147
B(t) = B(t,) x t
(, F, Px)
Figure 13.3: Continuous-time Brownian Motion, starting at x 6= 0.
IP (B(0) = 0) = 1:
For a Brownian motion B (t) that starts at x, denote the corresponding probability measure by IP x (See Fig. 13.3), and for such a Brownian motion we have:
IP x (B(0) = x) = 1:
Note that: If x 6= 0, then IP x puts all its probability on a completely different set from I . P The distribution of B (t) under IP x is the same as the distribution of x + B (t) under I . P
(s)-measurable
B{zs) |( }
3 7 ) F (s)7 5
148
B(s) s restart
Figure 13.4: Markov Property of Brownian Motion. Use the Independence Lemma. Dene
s+t
g (x) = IE h( B (s + t) ; B(s) + x )]
2 6 = IE 6h( x + 4
= IE xh(B (t)):
Then
B (t) |{z}
3 7 )7 5
= min ft 0 B(t) = xg :
Then we have:
IE h( B( + t) ) F ( ) = g(B( )) = IE x h(B(t)):
149
x restart
Figure 13.5: Strong Markov Property of Brownian Motion.
+t
(y
;x)2
2t
;1
;1
;1
= min ft 0
B (t) = xg :
> 0. Then
exp B (t ^ ) ; 1 2 (t ^ ) 2
is a martingale, and
1 IE exp B(t ^ ) ; 2 2 (t ^ ) = 1:
150 We have
(14.1)
IE expf x ; 1 2
Let
g1
<
1g
= 1:
#0 to get IE 1
<
1g
= 1, so
IP f < 1g = 1 IE expf; 1 2 g = e 2
Let
x:
(14.2)
=1 2
IEe
Differentiation of (14.3) w.r.t.
=e x
;
> 0:
p
(14.3)
yields
;IE e
Letting
= ; px e x 2
;
#0, we obtain
IE = 1:
(14.4)
Conclusion. Brownian motion reaches level x with probability 1. The expected time to reach level x is innite. We use the Reection Principle below (see Fig. 13.6).
IP f
t B (t) < xg = IP fB (t) > xg IP f tg = IP f t B(t) < xg + IP f t B (t) > xg = IP fB (t) > xg + IP fB (t) > xg = 2IP fB (t) > xg
Z e = p2 2 tx
1
y2 2t dy
151
shadow path
x t Brownian motion
Figure 13.6: Reection Principle in Brownian Motion. Using the substitution z
= yt
p
dz =
dy we get t
IP f
Density:
tg = p2
x pt
z2
2
dz:
x2 2t
@ f (t) = @t IP f F (t) =
tg = p x 3 e 2 t
Zb
a(t)
g (z) dz
then
@F = ; @a g (a(t)): @t @t IEe
;
= e t f (t)dt = e x
; ;
Z
0
152
Chapter 14
The It Integral o
The following chapters deal with Stochastic Differential Equations in Finance. References: 1. B. Oksendal, Stochastic Differential Equations, Springer-Verlag,1995 2. J. Hull, Options, Futures and other Derivative Securities, Prentice Hall, 1993.
B (0) = 0 Technically, IP f! B(0 !) = 0g = 1, B (t) is a continuous function of t, If 0 = t0 t1 : : : tn , then the increments B (t1) ; B (t0 ) : : : B(tn ) ; B(tn 1 )
;
FV (f ), of a
154
f(t)
t2 t1 T t
Figure 14.1: Example function f (t). For the function pictured in Fig. 14.1, the rst variation over the interval
0 T ] is given by:
Zt1
0 ZT 0
Zt2
ZT
t1
t2
= jf (t)j dt:
0
Thus, rst variation measures the total amount of up and down motion of the path. The general denition of rst variation is as follows: Denition 14.1 (First Variation) Let
0 = t0 t1 : : : tn = T:
The mesh of the partition is dened to be
We then dene
FV 0 T ](f ) = lim 0 !
jj jj
n 1 X
;
k=0
Suppose f is differentiable. Then the Mean Value Theorem implies that in each subinterval t k tk+1 ], there is a point tk such that
155
n 1 X
;
k=0
and
jf (tk+1) ; f (tk )j =
n 1 X
;
k=0
;
jf (tk )j(tk+1 ; tk )
0 0
FV 0 T ](f ) = lim 0 !
jj jj
n 1 X k=0
jf (tk )j(tk+1 ; tk )
= jf (t)j dt:
0
ZT
0
0 T]
n 1 X
;
Remark 14.1 (Quadratic Variation of Differentiable Functions) If f is differentiable, then hf i(T ) = 0, because
n 1 X
;
k=0
n 1 X
;
k=0
jf (tk )j2(tk+1 ; tk )2
0
jj jj:
and
n 1 X
;
k=0
jf (tk )j2(tk+1 ; tk )
0
hf i(T )
jj
n 1 X
;
= lim jj jj jf (t)j2 dt
jj
!0
ZT
0
k=0
jf (tk )j2(tk+1 ; tk )
0
= 0:
Theorem 3.44
hBi(T ) = T
or more precisely,
IP f! 2
hB(: !)i(T ) = T g = 1:
156 Proof: (Outline) Let = ft0 t1 : : : tn g be a partition of B (tk+1 ) ; B(tk ). Dene the sample quadratic variation
Q =
Then
n 1 X
;
k=0
2 Dk :
Q ;T =
We want to show that
jj jj
n 1 X
;
k=0
2 Dk ; (tk+1 ; tk )]:
lim (Q ; T ) = 0:
!0
IE (Q ; T ) = IE
For j
n 1 X
;
6= k, the terms
var(Q ; T ) = = =
n 1 X
;
k=0
2 Dk ; (tk+1 ; tk )] = 0: 2 Dk ; (tk+1 ; tk )
Dj2 ; (tj+1 ; tj )
and
are independent, so
k=0 n 1 X
;
k=0 n 1 X
;
k=0
=2
n 1 X
;
k=0
2jj jj
= 2jj jj T:
Thus we have
k=0
(tk+1 ; tk )
157
lim (Q ; T ) = 0:
jj
!0
n 1 X
;
hBi(T2) ; hBi(T1)] = T2 ; T1 = 1: T2 ; T1 T2 ; T1
1
As we increase the number of sample points, this approximation becomes exact. In other words, Brownian motion has absolute volatility 1. Furthermore, consider the equation
hBi(T ) = T = 1 dt
0
ZT
8T 0:
This says that quadratic variation for Brownian motion accumulates at rate 1 at all times along almost every path.
158
B(t) t
0, and the
s t=) every set in F (s) is also in F (t), B (t) is F (t)-measurable, 8t, For t t1 : : : tn , the increments B (t1 ) ; B (t) B (t2 ) ; B (t1 ) : : : B (tn ) ; B (tn 1 ) are independent of F (t).
;
The integrand is 1. 2.
(t) t 0, where
is adapted)
IE
We want to dene the It Integral: o
ZT
0
2(t) dt < 1
8T:
I (t) =
Zt
0
(u) dB (u)
Zt
0
(u) df (u) =
Zt
0
This wont work when the integrator is Brownian motion, because the paths of Brownian motion are not differentiable.
Assume that (t) is constant on each subinterval t k tk+1 ] (see Fig. 14.2). We call such a elementary process. The functions B (t) and
159
( t ) = ( t 1 ) ( t ) = ( t ) 0
( t )= ( t 3 )
0=t0
t1
t2
t3
t4 = T
( t ) = ( t 2 )
Figure 14.2: An elementary function . Think of t0 t1
Think of (tk ) as the number of shares of the asset acquired at trading date t k and held until trading date tk+1 . Then the It integral I (t) can be interpreted as the gain from trading at time t; this gain is given by: o
8 > (t0) B(t) ; | (t0) ] B{z } 0 t t1 > < =B (0)=0 I (t) = > (t0 ) B (t1) ; B (t0 )] + (t1 ) B (t) ; B (t1 )] t t t2 > : (t0) B(t1) ; B(t0)] + (t1) B(t2) ; B(t1 )] + (t2) B(t) ; B(t2)] t1 t t3: 2
t tk+1 , I (t) =
k 1 X
;
In general, if tk
j =0
Zt
0
(u) dB (u)
J (t) =
( (u)
Zt
0
(u) dB (u)
then
I (t) J (t) =
Zt
0
(u)) dB (u)
160
s t l t l+1 t
.....
k+1
cI (t) =
Martingale
Zt
0
c (u)dB (u):
I (t) is a martingale.
We prove the martingale property for the elementary process case. Theorem 7.45 (Martingale Property)
I (t) =
is a martingale.
k 1 X
;
j =0
tk t tk+1
s t be given. We treat the more difcult case that s and t are in different Proof: Let 0 subintervals, i.e., there are partition points t ` and tk such that s 2 t` t`+1 ] and t 2 tk tk+1 ] (See Fig. 14.3).
Write
I (t) =
` 1 X
;
j =0
j =`+1
2` 1 3 `1 X X IE 4 (tj )(B (tj+1 ) ; B(tj )) F (s)5 = (tj )(B (tj +1 ) ; B (tj )):
; ;
j =0
j =0
IE (t`)(B (t`+1 ) ; B(t` )) F (s) = (t` ) (IE B(t`+1)jF (s)] ; B(t` )) = (t` ) B (s) ; B (t` )]
161
2k 1 3 k1 X X IE 4 (tj )(B (tj +1 ) ; B (tj )) F (s)5 = IE IE (tj )(B (tj +1 ) ; B (tj )) F (tj ) F (s) j =`+1 j =`+1 3 2 k 1 X 6 = IE 4 (tj ) (IE B (tj +1)jF (tj )] ; B(tj )) F (s)7 5 {z } | j =`+1 =0 2 3 IE (tk )(B (t) ; B(tk )) F (s) = IE 6 (tk ) |IE B(t)jF ({z )] ; B(tk )) F (s)7 tk 4 ( 5 }
;
=0
IEI 2(t) = IE
Proof: To simplify notation, assume t = t k , so
Zt
0
2 (u) du:
I (t) =
k X
j =0
IEI 2(t) =
= =
IE 2(tj )Dj2] IE
2 (t
j )IE
IE 2(tj )(tj+1 ; tj )
2 (u) du
= IE
k X tZj+1
j =0 tj Zt 2 (u) du = IE 0
162
path of 4
path of
0=t0
t1
t2
t3
t4 = T
0 T ].
> 0. Let IE
(t) is F (t)-measurable, 8t 2 0 T ],
R T 2(t) dt < 1: 0
nlim IE !
1
ZT
0
j n(t) ; (t)j2 dt = 0:
In (T ) =
for every n. We now dene
ZT
0
n (t) dB (t)
ZT
0
ZT
0
n (t) dB (t):
163
The only difculty with this approach is that we need to make sure the above limit exists. Suppose n and m are large positive integers. Then
var(In (T ) ; Im (T )) = IE
(It Isometry:) = IE o
ZT
0 T
Z
0 0
!2
((a + b)2
j n(t) ; (t)j + j (t) ; m(t)j ]2 dt ZT ZT 2a2 + 2b2 :) 2IE j n (t) ; (t)j2 dt + 2IE j m (t) ; (t)j2 dt
= IE
0 0
1
ZT
which is small. This guarantees that the sequence fI n (T )gn=1 has a limit.
I (t) =
Zt
0
(u) dB (u):
Here is any adapted, square-integrable process. Adaptedness. For each t, I (t) is F (t)-measurable. Linearity. If
I (t) =
then
Zt
0
(u) dB (u)
J (t) =
( (u)
Zt
0
(u) dB (u)
I (t) J (t) =
Zt
0
(u)) dB (u)
and
cI (t) =
Zt
0
c (u)dB (u):
Martingale.
I (t) is a martingale. Continuity. I (t) is a continuous function of the upper limit of integration t. Rt It Isometry. IEI 2(t) = IE 0 2 (u) du. o
ZT
0
B(u) dB(u):
164
T/4
2T/4
3T/4
0 T ].
if if if
u < T:
By denition,
ZT
0
n 1
; X
B kT n k =0
B (k + 1)T ; B kT n n
so
We compute
1 2
n 1 k =0
; X
(Bk+1 ; Bk )2 = 1 2
; X
2 Bk+1 ;
n 1 j =0
n 1 k =0
; X
Bk Bk+1 + 1 2
n 1 k =0
n 1 k =0
; X
2 Bk
n 1 k =0
1 2 = 2 Bn + 1 2 1 2 = 2 Bn +
; X
Bj2 ;
; X
Bk Bk+1 + 1 2
; X
2 Bk
n 1 k =0 n 1
; X ; X
2 Bk ;
n 1 k =0
; X
Bk Bk+1
= Bn ;
1 2 2
k =0
Bk (Bk+1 ; Bk ):
165
n 1 k =0
; X
2 Bk (Bk+1 ; Bk ) = 1 Bn ; 1 2 2
n 1 k =0
; X
(Bk+1 ; Bk )2
n 1 k =0
or equivalently
n 1
; X
B kT n k =0
B (k + 1)T ; B kT n n
1 = 1 B 2 (T) ; 2 2
; X
B (k + 1)T n
k T
ZT
0
Remark 14.4 (Reason for the 1 T term) If f is differentiable with f (0) = 0, then 2
ZT
0
f (u) df (u) =
ZT
0
f (u)f (u) du
0
= 1 f 2 (u) 2
In contrast, for Brownian motion, we have
T
0
= 1 f 2 (T ): 2
ZT
0
B (u)dB(u) = 1 B 2(T ) ; 1 T: 2 2
The extra term 1 T comes from the nonzero quadratic variation of Brownian motion. It has to be 2 there, because Z
IE
B (u) dB (u) = 0
but
IE 1 B 2 (T ) = 1 T: 2 2
I (t) =
Then
Zt
0
(u) dB (u):
0 2(u) du:
hI i(t) =
Zt
166 This holds even if is not an elementary process. The quadratic variation formula says that at each time u, the instantaneous absolute volatility of I is 2 (u). This is the absolute volatility of the Brownian motion scaled by the size of the position (i.e. (t)) in the Brownian motion. Informally, we can write the quadratic variation formula in differential form as follows:
Proof: (For an elementary process ). Let = ft0 t1 : : : tn g be the partition for , i.e., (tk ) for tk t tk+1 . To simplify notation, assume t = tn . We have
hI i(t) =
Let us compute hI i(tk+1) ; hI i(tk ). Let
k=0
hI i(tk+1) ; hI i(tk)] :
tk = s0
Then
I (sj+1 ) ; I (sj ) =
so
(tk ) dB (u)
hI i(tk+1) ; hI i(tk) =
m 1 X
;
j =0
= 2 (tk )
jj jj!0 ;;;;;!
It follows that
j =0 2 (t
hI i(t) =
=
n 1 X
; ;
k=0 n 1 tZ +1 X k k=0 tk
jj 0 ; jj!! ;;;;;
Zt
Chapter 15
It s Formula o
15.1 It s formula for one Brownian motion o
We want a rule to differentiate expressions of the form f (B (t)), where f (x) is a differentiable function. If B (t) were also differentiable, then the ordinary chain rule would give
However, B (t) is not differentiable, and in particular has nonzero quadratic variation, so the correct formula has an extra term, namely,
dB (t) dB (t)
This is It s formula in differential form. Integrating this, we obtain It s formula in integral form: o o
f (B(t)) ; | (B (0)) = f {z }
f (0)
Zt
0
1 f (B(u)) dB (u) + 2
0
Zt
0
f (B (u)) du:
00
Remark 15.1 (Differential vs. Integral Forms) The mathematically meaningful form of It s foro mula is It s formula in integral form: o
f (B(t)) ; f (B (0)) =
Zt
0
1 (B (u)) dB (u) + 2 0
167
Zt
f (B (u)) du:
00
168 This is because we have solid denitions for both integrals appearing on the right-hand side. The rst,
Zt
0
f (B(u)) dB (u)
0
Zt
0
f (B (u)) du
00
is a Riemann integral, the type used in freshman calculus. For paper and pencil computations, the more convenient form of It s rule is It s formula in differo o ential form:
df (B (t)) = f (B (t)) dB(t) + 1 f (B (t)) dt: 2 There is an intuitive meaning but no solid denition for the terms df (B (t)) dB (t) and dt appearing
0 00
in this formula. This formula becomes mathematically respectable only after we integrate it.
f (x) = x f (x) = 1:
0 00 0 00
xk+1 be numbers. Taylors formula implies f (xk+1) ; f (xk ) = (xk+1 ; xk )f (xk ) + 1 (xk+1 ; xk )2f (xk ): 2
In the general case, the above equation is only approximate, and the error is of the order of (x k+1 ; xk )3 . The total error will have limit zero in the last step of the following argument. Fix T
In this case, Taylors formula to second order is exact because f is a quadratic function.
> 0 and let = ft0 t1 : : : tng be a partition of 0 T ]. Using Taylors formula, we write: f (B(T )) ; f (B(0)) 1 = 2 B 2 (T ) ; 1 B 2 (0) 2
= = =
n 1 X
; ;
k=0 n 1 X k=0 n 1 X
;
n 1 X
;
k=0
n 1 X
;
k=0
k=0
169
jj!0 to obtain
f (B (T )) ; f (B(0)) =
=
ZT ZT
0 0
T ZT (B (u)) dB (u) + 1 f 2 0 |
00
(B (u)) du: {z }
1
f (x) = 1 x2 : 2
15.3 Geometric Brownian motion
Denition 15.1 (Geometric Brownian Motion) Geometric Brownian motion is
1 ;2
;1 2
so
S (t) = f (t B (t)):
Then
ft =
According to It s formula, o
;1 2
f fx = f fxx = 2f:
dt
S (t) = S (0) +
Zt
0
S (u) du +
Zt
0
S (u) dB(u):
170
S (t) = S (0) +
the Riemann integral
Zt
0
S (u) du +
Zt
0
S (u) dB(u)
F (t) =
is differentiable with F 0 (t) =
Zt
0
S (t). This term has zero quadratic variation. The It integral o G(t) =
Zt
0
hGi(t) =
Zt
0
Thus the quadratic variation of S is given by the quadratic variation of G. In differential notation, we write
)]2
ZT2
2 S 2(u) du
As T2 # T1 , the above approximation becomes exact. In other words, the instantaneous relative volatility of S is 2 . This is usually called simply the volatility of S .
dX (t) = (t)dS (t) + r X (t) ; (t)S (t)] dt = (t) S (t)dt + S (t)dB (t)] + r X (t) ; (t)S (t)] dt = rX (t)dt + (t)S (t) ( {z r) dt + (t)S (t) dB (t): | ; }
Risk premium
Value of an option. Consider an European option which pays g (S (T )) at time T . Let v (t x) denote the value of this option at time t if the stock price is S (t) = x. In other words, the value of the option at each time t 2 0 T ] is
v (t S (t)):
dv (t S (t)) = vt dt + vxdS + 1 vxxdS dS 2 = vt dt + vx S dt + S dB ] + 1 vxx 2 S 2 dt h i 2 1 2 S 2v = vt + Svx + 2 xx dt + SvxdB A hedging portfolio starts with some initial wealth X 0 and invests so that the wealth X (t) at each time tracks v (t S (t)). We saw above that dX (t) = rX + ( ; r)S ] dt + S dB: To ensure that X (t) = v (t S (t)) for all t, we equate coefcients in their differentials. Equating the dB coefcients, we obtain the -hedging rule: (t) = vx (t S (t)): Equating the dt coefcients, we obtain: vt + Svx + 1 2 S 2vxx = rX + ( ; r)S: 2 But we have set = vx , and we are seeking to cause X to agree with v . Making these substitutions,
1 vt + Svx + 2 2S 2vxx = rv + vx ( ; r)S (where v = v (t S (t)) and S = S (t)) which simplies to vt + rSvx + 1 2 S 2vxx = rv: 2 In conclusion, we should let v be the solution to the Black-Scholes partial differential equation 1 vt (t x) + rxvx (t x) + 2 2 x2vxx(t x) = rv (t x)
satisfying the terminal condition If an investor starts with X 0 = v (0 S (0)) and uses the hedge (t) = vx (t X (t) = v(t S (t)) for all t, and in particular, X (T ) = g (S (T )). we obtain
v (T x) = g (x):
172
r(t) dB(t)
bc
0 0 2(t). This is df (r(t)), where f (x) = x2 . We obtain We apply It s formula to compute dr o dr2(t) = df (r(t)) = f (r(t)) dr(t) + 1 f (r(t)) dr(t) dr(t) 2
0 00
Zt
Z tq
r(u) dB (u):
r(t) dB (t)
= 2abr(t) dt ; 2acr2(t) dt + 2 r 2 (t) dB (t) + 2 r(t) dt 3 = (2ab + 2 )r(t) dt ; 2acr2(t) dt + 2 r 2 (t) dB (t)
The mean of r(t). The integral form of the CIR equation is
Zt
Z tq
0
r(u) dB (u):
Taking expectations and remembering that the expectation of an It integral is zero, we obtain o
Zt
Zt
0
;
b If r(0) = c , then IEr(t) = b for every t. If r(0) 6= b , then r(t) exhibits mean reversion: c c
IEr(t) = b + e c
act
lim IEr(t) = b : t!
1
173
Zt
0
r(u) du ; 2ac
Zt
0
r2(u) du + 2
Zt
0
2 r 3 (u) dB(u):
Zt
0
IEr(u) du ; 2ac
Zt
0
IEr2(u) du:
d e2actIEr2(t) = e2act 2acIEr2(t) + d IEr2(t) dt dt 2act (2ab + 2)IEr(t): =e Using the formula already derived for IEr(t) and integrating the last equation, after considerable
algebra we obtain
2 2b b 2 + b2 + r(0) ; b act 2 c2 2ac c ac + c e 2 2 2 + r(0) ; b e 2act + ac 2bc ; r(0) e 2act : c ac var r(t) = IEr2(t) ; (IEr(t))2 2 2 b 2 = 2ac2 + r(0) ; b ac e act + ac 2bc ; r(0) e 2act : c
IEr2(t) =
with the following properties: Each Bk (t) is a one-dimensional Brownian motion; If i 6= j , then the processes Bi (t) and Bj (t) are independent. Associated with a d-dimensional Brownian motion, we have a ltration fF (t)g such that For each t, the random vector B (t) is F (t)-measurable; For each t
t1 : : : tn , the vector increments B(t1 ) ; B(t) : : : B(tn) ; B(tn 1 ) are independent of F (t).
;
174
C =
n 1 X
;
k=0
The increments appearing on the right-hand side of the above equation are all independent of one another and all have mean zero. Therefore,
IEC = 0:
We compute var(C ). First note that
C2 =
+2
n 1 X
;
k=0 n 1 X
;
`<k
All the increments appearing in the sum of cross terms are independent of one another and have mean zero. Therefore,
var(C ) = IEC 2 = IE
n 1 X
;
k=0
But Bi (tk+1 ) ; Bi (tk )]2 and Bj (tk+1 ) ; Bj (tk )]2 are independent of one another, and each has expectation (tk+1 ; tk ). It follows that
var(C ) =
As jj
n 1 X
;
k=0
(tk+1 ; tk )2 jj jj
n 1 X
;
k=0
(tk+1 ; tk ) = jj jj:T: = 0.
175
Zt
0
Z0t
(u) du + (u) du +
Zt Z0t
0
Zt Z0t
0
Such processes, consisting of a nonrandom initial condition, plus a Riemann integral, plus one or more It integrals, are called semimartingales. The integrands (u) (u) and ij (u) can be any o adapted processes. The adaptedness of the integrands guarantees that X and Y are also adapted. In differential notation, we write
dX = dt + 11 dB1 + 12 dB2 dY = dt + 21 dB1 + 22 dB2 : Given these two semimartingales X and Y , the quadratic and cross variations are: dX dX = ( dt + 11 dB1 + 12 dB2)2 2 2 = 11 dB1{z 1 +2 11 12 dB1{z 2 + 12 dB2{z 2 | dB } | dB} | dB}
=( dY dY = ( =( dX dY = ( =(
0 dt 2 + 2 )2 dt 11 12 dt + 21 dB1 + 22 dB2 )2 2 2 2 21 + 22 ) dt dt + 11 dB1 + 12 dB2)( 11 21 + 12 22 ) dt
dt
dt +
21 dB1 + 22 dB2 )
Let f (t x y ) be a function of three variables, and let X (t) and Y (t) be semimartingales. Then we have the corresponding It formula: o
df (t x y) = ft dt + fx dX + fy dY + 1 fxx dX dX + 2fxy dX dY + fyy dY dY ] : 2 In integral form, with X and Y as decribed earlier and with all the variables lled in, this equation
is
0Z
11 fx + 21 fy ]
dB1 +
Zt
0
2 1 2 11 21 + 12 22)fxy + 2 ( 21 + 22)fyy ] du
12 fx + 22 fy ]
dB2
176
Chapter 16
(SDE)
x) and Lips-
j (t x) ; (t y)j Ljx ; yj
t0
x y. Let (t0 x) be given. A solution to (SDE) with the initial condition (t 0 x) is a process fX (t)gt X (t0) = x
Zt
Zt
t0
(s X (s)) dB (s)
t t0
The solution process fX (t)g t t0 will be adapted to the ltration fF (t)g t 0 generated by the Brownian motion. If you know the path of the Brownian motion up to time t, then you can evaluate X (t).
Example 16.1 (Drifted Brownian motion) Let a be a constant and
= 1, so
dX(t) = a dt + dB(t):
If (t0
178
then
X(t) = x + a(t ; t0) + (B(t) ; B(t0 )) To compute the differential w.r.t. t, treat t0 and B(t0 ) as constants: dX(t) = a dt + dB(t):
t t0:
be constants. Consider
X(t0 ) = x
the solution is
X(t) = x exp (B(t) ; B(t0 )) + (r ; 1 2 )(t ; t0) : 2 Again, to compute the differential w.r.t. t, treat t0 and B(t0 ) as constants: dX(t) = (r ; 1 2)X(t) dt + X(t) dB(t) + 1 2 X(t) dt 2 2 = rX(t) dt + X(t) dB(t):
X (0) = :
179
p(t0 t1 x y) the density (in the y variable) of X (t 1), conditioned on X (t 0) = x. In other words, IE t0 xh(X (t1)) =
The Markov property says that for 0
IR
IE 0
IR
dX(t) = a dt + dB(t):
Conditioned on (t1 ; t0 ), i.e.,
X(t 0 ) = x, the random variable X(t1 ) is normal with mean x + a(t1 ; t0 ) and variance
2 p(t0 t1 x y) = p 1 exp ; (y ; (x + a(t1 ; t0 ))) : 2(t1 ; t0 ) 2 (t1 ; t0) Note that p depends on t0 and t1 only through their difference t 1 ; t0 . This is always the case when a(t x) and (t x) dont depend on t.
Example 16.4 (Geometric Brownian motion) Recall that the solution to the SDE
y = x exp b + (r ; 1 2)(t1 ; t0 ) 2
or equivalently,
The derivative is
db = dy : y
180
Therefore,
Using the transition density and a fair amount of calculus, one can compute the expected payoff from a European call:
+
Z1
where
N( ) = p1 2
Therefore,
Z1 12 1 2 e; 2 x dx = p1 e; 2 x dx: 2 ; ;1
IE 0 e;r(T ;t) (X(T) ; K)+ F (t) = e;r(T ;t) IE t X (t) (X(T) ; K)+ 1 = X(t)N p 1 log X(t) + r(T ; t) + 2 2(T ; t) K T ;t log X(t) + r(T ; t) ; 1 2 (T ; t) ; e;r(T ;t) K N p 1 2 K T ;t
x y ) be the transition density. Then the Kolmogorov Backward Equation is: @ @ @2 1 ; @t p(t0 t1 x y) = a(t0 x) @x p(t0 t1 x y) + 2 2(t0 x) @x2 p(t0 t1 x y): 0 (KBE)
In the case that a and are functions of x alone, p(t0 t1 x y ) depends on t0 and t1 only through their difference = t1 ; t0 . We then write p( x y ) rather than p(t0 t1 x y ), and (KBE ) becomes
(KBE)
181
Therefore,
2 1 apx + 2 pxx = a(y ; x ; a ) ; 21 + (y ; x ; a ) p 2 2 =p :
This is the Kolmogorov backward equation. Example 16.6 (Geometric Brownian motion)
dX(t) = rX(t) dt + X(t) dB(t): h y 1 1 p( x y) = p exp ; 2 1 2 log x ; (r ; 2 2 ) y 2 It is true but very tedious to verify that p satises the KBE 1 p = rxpx + 2 2x2 pxx:
i2
(5.1)
v (t x) = IE t xh(X (T ))
182 where 0
t T . Then
v(t x) = h(y) p(T ; t x y ) dy vt(t x) = ; h(y ) p (T ; t x y ) dy vx(t x) = h(y) px(T ; t x y ) dy vxx(t x) = h(y) pxx(T ; t x y ) dy:
Therefore, the Kolmogorov backward equation implies
Z Z Z
) be an initial condition for the SDE (5.1). We simplify notation by writing IE rather than
It s formula implies o
183
Z0t
0
Rt We know that v (t X (t)) is a martingale, so the integral 0 for all t. This implies that the integrand is zero; hence
vt + avx + 1 2vxx = 0: 2
Thus by two different arguments, one based on the Kolmogorov backward equation, and the other based on It s formula, we have come to the same conclusion. o Theorem 5.51 (Feynman-Kac) Dene
v(t x) = IE t xh(X (T ))
where Then
0 t T
and
v (T x) = h(x):
The Black-Scholes equation is a special case of this theorem, as we show in the next section. Remark 16.1 (Derivation of KBE) We plunked down the Kolmogorov backward equation without any justication. In fact, one can use It s formula to prove the Feynman-Kac Theorem, and use o the Feynman-Kac Theorem to derive the Kolmogorov backward equation.
16.6 Black-Scholes
Consider the SDE With initial condition the solution is
u t:
184 Dene
IE h(X Y ) G = (X )
where
(x) = IEh(x Y ):
With geometric Brownian motion, for 0
S (t) = S (0) exp B(t) + (r ; 1 2)t 2 n o S (T ) = S (0) exp B(T ) + (r ; 1 2)T 2 n o 1 = S (t) exp (B (T ) ; B (t)) + (r ; 2 2 )(T ; t) |{z}
F
t T , we have
(t)-measurable
independent of F (t)
{z
We thus have
S (T ) = XY
where
185
0 t T:
Note that the random variable h(S (T )) whose conditional expectation is being computed does not depend on t. Because of this, the tower property implies that v (t S (t)) 0 t T , is a martingale: For 0 s t T ,
v(t S (t)) is a martingale, the sum of the dt terms in dv (t S (t)) must be 0. dv (t S (t)) = vt(t S (t)) dt + rS (t)vx(t S (t)) + 1 2 S 2(t)vxx(t S (t)) dt 2 + S (t)vx(t S (t)) dB (t):
By It s o
0 t < T x 0:
Along with the above partial differential equation, we have the terminal condition
Finally, we shall eventually see that the value at time t of a contingent claim paying h(S (T )) is
v(t 0) = h(0)
; ;
0 t T:
u(t x) = e =e
at time t if S (t) = x. Therefore,
x) = er(T t)u(t x) x) = ;rer(T t)u(t x) + er(T t)ut (t x) x) = er(T t)ux(t x) x) = er(T t)uxx(t x):
; ; ; ; ;
186 Plugging these formulas into the partial differential equation for v and cancelling the e r(T ;t) appearing in every term, we obtain the Black-Scholes partial differential equation:
0 t < T x 0:
(BS)
Compare this with the earlier derivation of the Black-Scholes PDE in Section 15.6.
y 1 log x ; (r ; 2 2)(T ; t)
2)
for geometric Brownian motion (See Example 16.4), we have the stochastic representation
u(t x) = e =e
In the case of a call, and
r(T t) IE t xh(S (T ))
; ;
r(T t)
(SR)
h(y ) = (y ; K )+
u(t x) = x N
p1
T ;t Even if h(y ) is some other function (e.g., h(y ) = (K ; y )+ , a put), u(t x) is still given by and
16.7 Black-Scholes with price-dependent volatility
0 t < T x > 0:
187
v(t 0) = 0
0 t T:
An example of such a process is the following from J.C. Cox, Notes on options pricing I: Constant elasticity of variance diffusions, Working Paper, Stanford University, 1975:
0 t<T x>0 v (t 0) = 0 0 t T v (T x) = (x ; K )+ x 0:
188
Chapter 17
e B (t) =
Zt
0
(u) du + B (t)
Z (t) = exp ;
Zt
0
(u) dB (u) ;
Zt 2 1 2 0 (u) du 8A 2 F :
Z
A
Z (T ) dIP
0 t T , is a Brownian motion.
IE exp
everything is OK. We make the following remarks:
( ZT
1 2 0
2 (u) du
<1
Z (t) is a matingale. In fact, 1 dZ (t) = ; (t)Z (t) dB (t) + 2 2 (t)Z (t) dB(t) dB(t) ; 1 2(t)Z (t) dt 2 = ; (t)Z (t) dB (t):
189
190
f IP is a probability measure.
f IP ( ) =
0. In particular
Z (T ) dIP = IEZ (T ) = 1
f IE in terms of IE .
f so I is a probability measure. P
f IEZ = IE Z (T )X ] : To see this, consider rst the case X = 1 A , where A 2 F . We have Z Z f f IEX = IP (A) = Z (T ) dIP = Z (T )1A dIP = IE Z (T )X ] :
A
Now use Williams standard machine. The intuition behind the formula
f IP and IP .
f IP (A) =
Z
A
Z (T ) dIP
8A 2 F f IP . Thus,
e Distribution of B (T ). If
f but since IP (! ) = 0 and I (!) = 0, this doesnt really tell us anything useful about P we consider subsets of , rather than individual elements of .
is constant, then
Z (T ) = exp ; B (T ) ; 1 2 T 2 e (T ) = T + B(T ): B e Under IP , B (T ) is normal with mean 0 and variance T , so B (T ) is normal with mean T and variance T : ( ) b 1 exp ; (~ ; T )2 d~: e IP (B (T ) 2 d~) = p b b 2T 2 T e f e P Removal of Drift from B (T ). The change of measure from IP to I removes the drift from B (T ).
To see this, we compute
Z = p1 2 T 1 Z =p 2 T Z (y = T + b) = p 1 2 T = 0:
( T + b) expf; b ;
;1 1
( T + b) exp ; (b + TT ) 2
2 y exp ; y2 dy
1 2 2 Tg
( 2) b exp ; 2T db 2)
db
= T + b)
;1
(Substitute y
191
fe E We can also see that I B (T ) = 0 by arguing directly from the density formula
2 ~ IP B (t) 2 db = p 1 exp ; (b ; TT ) d~: b 2 2 T
Because
ne
o ~
f e e IP B(T ) 2 d~ = IP B (T ) 2 d~ exp ; ~ + 1 2 T b b b 2 ( ~ ) 1 exp ; (b ; T )2 ; ~ + 1 2 T d~: b 2 =p b 2T 2 T ( ) b 1 exp ; ~2 d~: =p 2T b 2 T f e Under I , B (T ) is normal with mean zero and variance P mean T and variance T .
T.
Under IP ,
e B (T ) is normal with
Means change, variances dont. When we use the Girsanov Theorem to change the probability measure, means change but variances do not. Martingales may be destroyed or created. Volatilities, quadratic variations and cross variations are unaffected. Check:
t T . If X is F (t)-measurable, then
Proof:
0 t T , is a martingale under IP .
s t T , then
(1.1)
Proof: It is clear that Z 1s) IE XZ (t)jF (s)] is ( property. For A 2 F (s), we have
F (s)-measurable.
1 IE XZ (t)jF (s)] dIP = IE 1 1 IE XZ (t)jF (s)] f f A Z (s) A Z (s) (Lemma 1.53) = IE 1A IE XZ (t)jF (s)]] = IE IE 1A XZ (t)jF (s)]] (Taking in what is known) = IE 1A XZ (t)] f = IE 1A X ] (Lemma 1.53 again) Z f = X dIP :
A
Although we have proved Lemmas 1.53 and 1.54, we have not proved Girsanovs Theorem. We will not prove it completely, but here is the beginning of the proof. Lemma 1.55 Using the notation of Girsanovs Theorem, we have the martingale property
0 s t T:
e Proof: We rst check that B (t)Z (t) is a martingale under IP . Recall e dB (t) = (t) dt + dB (t) dZ (t) = ; (t)Z (t) dB(t):
Therefore,
fe e IE B (t)jF (s)] = Z 1s) IE B (t)Z (t)jF (s)] ( e = Z 1s) B (s)Z (s) ( e = B (s):
193
Denition 17.1 (Equivalent measures) Two measures on the same probability space which have the same measure-zero sets are said to be equivalent. The probability measures dened by
IP
and
Recall that
f IP is
Z f(A) = Z (T ) dIP IP Z
dIP = 0: Because Z (T ) > 0 for every !, we can invert the denition IP (A) = dIP A Z (T )
1
A 2 F:
dIP = 0:
T,
r(t) 0 t T . The process r(t) is adapted. Wealth of an agent, starting with X (0) = x. We can write the wealth process differential in
Interest rate: several ways:
dX (t) =
= r(t)X (t) dt + (t) dS (t) ; rS (t) dt] = r(t)X (t) dt + (t) ( (t) {z r(t)) S (t) dt + (t) (t)S (t) dB (t) | ; }
(t)
d e d e
R t r(u) du
0
R t r(u) du
0
S (t) = e
X (t) = e
R t r(u) du
0
;
R t r(u) du
0
= (t)d e
Notation:
R t r(u) du
Rt (t) = e 0 r(u) du
d (t) = r(t) (t) dt
= 1t) (t)S (t) (t) dt + dB (t)] ( (t) d X((tt)) = (t) d S(t) () = (tt) (t)S (t) (t) dt + dB (t)]:
Changing the measure. Dene
(t) d S(t) = 1t) ;r(t)S (t) dt + dS (t)] ( = 1t) ( (t) ; r(t))S (t) dt + (t)S (t) dB (t)] (
e B (t) =
Then
Zt
0
(u) du + B (t):
(t) e d S(t) = 1t) (t)S (t) dB (t) ( e d X((tt)) = ((tt)) (t)S (t) dB (t):
195
f IP (A) =
where
Z
A
0
Z (T ) dIP
(u) dB (u) ; 1 2
A2F
Z (t) = exp ;
Zt
Zt
(t) =
(t) 6=
Risk-neutral valuation. Consider a contingent claim paying an F (T )-measurable random variable V at time T .
Example 17.1
V = (S(T) ; K)+ European call + V = (K ; S(T)) European put !+ ZT 1 S(u) du ; K V= T Asian call 0 V = 0maxT S(t) Look back t
196
Chapter 18
X (t) = X (0) +
Zt
0
(u) dB (u)
0 t T:
In particular, the paths of X are continuous. Remark 18.1 We already know that if X (t) is a process satisfying
dX (t) = (t) dB (t) then X (t) is a martingale. Now we see that if X (t) is a martingale adapted to the ltration generated by the Brownian motion B (t), i.e, the Brownian motion is the only source of randomness in X , then dX (t) = (t) dB(t) for some (t).
18.2 A hedging application
Homework Problem 4.5. In the context of Girsanovs Theorem, suppse that F (t) the ltration generated by the Brownian motion B (under IP ). Suppose that Y is a Then there is an adapted process (t) 0 t T , such that
0 t T is f IP -martingale.
Y (t) = Y (0) +
Zt
0
e (u) dB (u)
197
0 t T:
198
dS (t) = (t)S (t) dt + (t)S (t) dB (t) Zt (t) = exp r(u) du 0 (t) = (t) ; r(t) Z t (t) e B (t) = (u) du + B(t)
0 Z f(A) = Z (T ) dIP IP
Z (t) = exp ;
A
Zt
(u) dB (u) ; 1 2
Zt
0
2 (u) du
8A 2 F :
Then
Let
(t) (t) e d S(t) = S(t) (t) dB (t): (t) 0 t T be a portfolio process. The corresponding wealth process X (t) satises (t) e d X((tt)) = (t) (t) S(t) dB(t)
i.e.,
0 t T:
Let V be an F (T )-measurable random variable, representing the payoff of a contingent claim at time T . We want to choose X (0) and (t) 0 t T , so that
X (T ) = V:
f V Y (t) = IE (T ) F (t) Zt
0
Y (t) = Y (0) +
V (T ) and choose
e (u) dB (u)
(u) so that
199
0 t T:
In particular,
X (T ) = IE V F (T ) = V f (T ) (T ) (T )
so
X (T ) = V:
The Martingale Representation Theorem guarantees the existence of a hedging portfolio, although it does not tell us how to compute it. It also justies the risk-neutral pricing formula
0 t T
Zt
0
(u) dB (u) ;
Zt
0
(r(u) + 1 2 (u)) du 2
F (t) 0 t T
t T
dene
For 0
F P);
T , a d-
e Bj (t) =
Zt
0
Z f(A) = Z (T ) dIP: IP
A
Z (t) = exp ;
j (u) du + Bj (t) Zt
j = 1 ::: d
1 (u): dB (u) ; 2 0 0
Zt
jj (u)jj2 du
200
0 t T
t T , is a martingale (under IP ) relative to F (t) 0 If X (t) 0 d-dimensional adpated process (t) = ( 1(t) : : : d (t)), such that X (t) = X (0) +
F (t) 0 t T
P);
Zt
0
(u): dB (u)
0 t T:
Corollary 4.59 If we have a d-dimensional adapted process (t) = ( 1 (t) : : : d (t)) then we can e f f P P dene B Z and I as in Girsanovs Theorem. If Y (t) 0 t T , is a martingale under I relative to F (t) 0 t T , then there is a d-dimensional adpated process (t) = ( 1(t) : : : d(t)) such that Z
Y (t) = Y (0) +
e (u): dB (u)
0 t T:
following:
B (t) = (B1(t) : : : Bd(t)) 0 t T , be a d-dimensional Brownian motion on some ( F P), and let F (t) 0 t T , be the ltration generated by B . Then we can dene the
Let
Stocks
d X j =1
i = 1 ::: m
(t) = exp
Here, i (t)
Zt
0
r(u) du :
201
d X d Si((tt)) = ( i(t){z r(t)) Si((tt)) dt + Si((tt)) ; } ij (t) dBj (t) | j =1 d ? Si (t) X (t) (t) + dB (t)] = (t) ij | j {z j } j =1 e dBj (t)
Risk Premium
(5.1)
:::
d (t), so that
(MPR)
d X j =1
i = 1 : : : m: :::
Market price of risk. The market price of risk is an adapted process (t) = ( 1 (t) satisfying the system of equations (MPR) above. There are three cases to consider:
d (t))
Case I: (Unique Solution). For Lebesgue-almost every t and IP -almost every ! , (MPR) has a unique solution (t). Using (t) in the d-dimensional Girsanov Theorem, we dene a unique f f P P risk-neutral probability measure I . Under I , every discounted stock price is a martingale. f Consequently, the discounted wealth process corresponding to any portfolio process is a I P martingale, and this implies that the market admits no arbitrage. Finally, the Martingale Representation Theorem can be used to show that every contingent claim can be hedged; the market is said to be complete. Case II: (No solution.) If (MPR) has no solution, then there is no risk-neutral probability measure and the market admits arbitrage. Case III: (Multiple solutions). If (MPR) has multiple solutions, then there are multiple risk-neutral probability measures. The market admits no arbitrage, but there are contingent claims which cannot be hedged; the market is said to be incomplete. Theorem 5.60 (Fundamental Theorem of Asset Pricing) Part I. (Harrison and Pliska, Martingales and Stochastic integrals in the theory of continuous trading, Stochastic Proc. and Applications 11 (1981), pp 215-260.): If a market has a risk-neutral probability measure, then it admits no arbitrage. Part II. (Harrison and Pliska, A stochastic calculus model of continuous trading: complete markets, Stochastic Proc. and Applications 15 (1983), pp 313-316): The risk-neutral measure is unique if and only if every contingent claim can be hedged.
202
Chapter 19
In what follows, all processes can depend on t and ! , but are adapted to simplify notation, we omit the arguments whenever there is no ambiguity. Stocks:
F (t) 0 t T
t T
F P). Let
t T.
To
F (t) 0
dS1 = S1 dS2 = S2
We assume 1
dt + 1 dB1] q 2 dt + 2 dB1 + 1 ;
1: Note that
2 2 dB2
>0
2>0
;1
2 2 dS1 dS2 = S1 2 dB1 dB1 = 2S1 dt 1 1 2 2 dS2 dS2 = S2 2 2 dB1 dB1 + S2 (1 ; 2) 2 dB2 dB2 2 2 2 S 2 dt = 2 2 dS1 dS2 = S1 1S2 2 dB1 dB1 = 1 2S1S2 dt:
In other words,
dS1 has instantaneous variance 2 , 1 S1 dS2 has instantaneous variance 2 , 2 S2 dS1 and dS2 have instantaneous covariance S1 S2
Accumulation factor:
1 2.
(t) = exp
Zt
0
r du :
= 2=
1;r
(MPR)
2 1+
1;
1 1 2 2
2;r
203
1 2
provided ;1 <
= =
1;r 1 ( 2 ; r) p 2( 1 ; r) ; 1 1 2 1; 2
Z f(A) = Z (T ) dIP IP
A
Z (t) = exp
8A 2 F :
Zt Z0
0
1 du + B1 (t) 2 du + B2 (t):
dS1 = S1 r dt + dS2 = S2 r dt +
e 1 dB1
i q
2 2 dB2 e
e 2 dB1 + 1 ;
We have changed the mean rates of return of the stock prices, but not the variances and covariances.
q e1 + 1 ; 2 dB
0 t T:
2 2
e dB2 :
205
Y (t) = Y (0) +
We have
Zt
0
e 1 dB1 +
Zt
0
e 2 dB2: e dB1
d X = 1 1 S1 1 + 1 2 S2 dY =
We solve the equations
e e 1 dB1 + 2 dB2 :
1 S1 1 +
+1
2S2
1;
2 2 dB2 e
1
for the hedging portfolio ( 1
2S2 1 ;
2 S2
2= 1 2 2= 2 2) and setting
f V X (0) = Y (0) = IE (T )
we have X (t) = Y (t)
0 t T
and in particular,
X (T ) = V:
Every F (T )-measurable random variable can be hedged; the market is complete.
=1
dS1 = S1 dS2 = S2
The stocks are perfectly correlated. The market price of risk equations are
1 dt + 1 dB1 ] 2 dt + 2 dB1 ]
1 1= 1;r 2 1= 2;r
The process 2 is free. There are two cases:
(MPR)
206 Case I: 1 ;r 6= 2 ;r : There is no solution to (MPR), and consequently, there is no risk-neutral 1 2 measure. This market admits arbitrage. Indeed
1S1
( 1 ; r) dt + 1
dB1] + 1 2S2 ( 2 ; r) dt +
1 : S
2 ; r dt + dB 1 2
dB1]
>
2; 2
r : Set
1= 1 S1 2=; 2 2
Then
d X =1
=1
1; 1
1 ; r dt + dB ; 1 1 1 1 ; r ; 2 ; r dt 1 {z 2 }
Positive
Case II:
2; 2
= =
1;r 2;r
1=
f P 2 is free; there are innitely many risk-neutral measures. Let I be one of them.
Hedging:
1;r 1
2;r 2
d X = 1 1S1 ( 1 ; r) dt +
=1 = 1
1 S1 1 +
dB1] + 1 2S2 ( 2 ; r) dt +
1
2 S2 2 1 dt + dB1 ]
dB1]
1S1 1 1 dt + dB1 ] +
2S2 2
e dB1 :
Let V be an F (T )-measurable random variable. If V depends on B2 , then it can probably not be hedged. For example, if
V = h(S1 (T ) S2(T ))
207
0 t T:
Y (t) = Y (0) +
so
Zt
0
e 1 dB1 +
Zt
0
e 2 dB2
dY =
e e 1 dB1 + 2 dB2 :
2 = 0:
208
Chapter 20
M (T ) = 0maxT B (t): t
Then we have:
= IP fB (T ) > 2m ; bg ( 2) 1 Z x =p exp ; 2T dx 2 T 2m b
1 ;
m>0 b<m
Z 2 @2 IP fM (T ) 2 dm B(T ) 2 dbg = ; @m @b p 1 exp ; x dx dm db 2T 2 T 2m b ( )! @ p 1 exp ; (2m ; b)2 dm db = ; @m 2T 2 T ( ) 2(2p ; b) exp ; (2m ; b)2 dm db m > 0 b < m: m = 2T T 2 T
1 ;
) !
e B(t) = t + B(t)
209
210
2m-b
shadow path
m b Brownian motion
Figure 20.1: Reection Principle for Brownian motion without drift
M (T ).
211
e B (T ):
Let h(m ~
f f e e = IE h(M (T ) B (T )) expf B (T ) ; 1 2 T g 2
=
m= ~Z
1
~=m bZ ~
;1
m=0 ~= ~ b
But also,
f e IEh(M (T ) B (T )) =
m= ~Z
~Z m b= ~
;1
m=0 ~= ~ b
f h(m ~) IP fM (T ) 2 dm B (T ) 2 d~g: ~ b ~ e b
f IP fM (T ) 2 dm B (T ) 2 d~g ~ e b 1 2 T g I fM (T ) 2 dm B (T ) 2 d~g f f e = expf ~ ; 2 b b )~ (P m ~ ~ ~ ~2 b 2 = 2(2p ; b) exp ; (2m ; b) : expf ~ ; 1 2 T gdm d~ m > 0 ~ < m: ~ b ~ b ~ 2T T 2 T
212
To simplify notation, assume that IP is already the risk-neutral measure, so the value at time zero of the option is
S (T ) = 0maxT S (t): t
v (0 S (0)) = e rT IE (S (T ) ; K )+ 1 S (T )<L : Because IP is the risk-neutral measure, dS (t) = rS (t) dt + S (t) dB (t) S (t) = S0 expf B (t) + (r ; 1 2 )tg 2
; f g
e = S0 expf B (t)g
2 6 6B(t) + r ; 6 4 | {z 2
where
v (0 S (0)) = e
rT IE
i
+
e = e rT IE S (0) expf B (T )g ; K 1
;
"
1 S(0)exp M (T ) e
f f
<L
e B (T )>
m ~
213
~ b
f M (T ).
S (0) K < L so 0 ~ < m: b ~ The other case, K < S (0) L leads to ~ < 0 m and the analysis is similar. b ~ R m R m : : :dy dx: ~ ~ We compute ~ x b
;
; 2 1 p v (0 S (0)) = e (S (0) expf xg ; K ) 2(2y ; x) exp ; (2y2T x) + x ; 2 2 T dy dx ~ x b T 2 T( ) ~ Zm ~ 1 exp ; (2y ; x)2 + x ; 1 2 T y=m dx rT = ;e (S (0) expf xg ; K ) p 2 ~ 2T 2 T b y=x " ( 2 ) Zm ~ = e rT ~ (S (0) expf xg ; K ) p 1 exp ; x + x ; 1 2 T 2 2T b 2 T )# ( (2m ; x)2 + x ; 1 2 T dx ~ ; exp ; 2T 2 ( ) ~ 1 e rT S (0) Z m exp x ; x2 + x ; 1 2 T dx =p 2 ~ 2T 2 T b ) ( ~ 1 e rT K Z m exp ; x2 + x ; 1 2 T dx ; p2 T 2 ~ 2T b ( ) ~ 1 e rT S (0) Z m exp x ; (2m ; x)2 + x ; 1 2 T dx ~ ;p 2 ~ 2T 2 T b ) ( ~ 1 e rT K Z m exp ; (2m ; x)2 + x ; 1 2 T dx: ~ +p 2 ~ 2T 2 T b
rT
; ; ; ; ; ;
Z mZ m ~ ~
The standard method for all these integrals is to complete the square in the exponent and then recognize a cumulative normal distribution. We carry out the details for the rst integral and just
214 give the result for the other three. The exponent in the rst integrand is
= ; 21 (x ; T ; T )2 + 1 2T + T 2 T 2 T = ; 21 x ; rT ; 2 + rT: T
In the rst integral we make the change of variable
x2 x ; 2T + x ; 1 2 T 2
rT ~ e p S (0) Z m exp b 2 T ~
;
x2 x ; 2T + x ; 1 2T dx 2 ( ) ~ 1 S (0) Z m exp ; 1 x ; rT ; T 2 dx =p ~ 2T 2 2 T b
p m p p~T ; r T ; 2 T
= p 1 S (0): 2 T ~ b
~ m ~ = S (0) N p ; r T ; 2 T ; N pb ; r T ; 2 T T T
Putting all four integrals together, we have
"
p p pT ; r T ; 2 T
expf; y2 g dy
p !
p !#
"
p !
p !#
where
~ = 1 log K b S (0)
m = 1 log SL : ~ (0)
215
v(t,L) = 0
v(T,x) = (x - K)+
v(t,0) = 0
Figure 20.4: Initial and boundary conditions. If we let L!1 we obtain the classical Black-Scholes formula
"
~ b
p !#
If we replace T by T ; t and replace S (0) by x in the formula for v (0 S (0)), we obtain a formula for v (t x), the value of the option at the time t if S (t) = x. We have actually derived the formula under the assumption x K L, but a similar albeit longer formula can also be derived for K < x L. We consider the function
v(t x) = IE t x e
r(T t) (S (T ) ; K )+ 1
;
S (T )<L
0 t T 0 x L:
v(T x) = (x ; K )+
and the boundary conditions
0 x<L
v(t 0) = 0 0 t T v(t L) = 0 0 t T:
We show that v satises the Black-Scholes equation
216 Let S (0) > 0 be given and dene the stopping time
e
is a martingale. Proof: First note that
r(t )v (t ^
^
S (t ^ )) 0 t T
rT (S (T ) ; K )+ 1
S (T )<L
F (t) (!) = 0:
(!) t, we have
IE e
rT (S (T ) ; K )+ 1
S (T )<L
F (t) (!) = e
r(t (!)) v (t ^
^
(!) S (t ^ (! ) !)) :
IE e
h t S (t !)
rT (S (T ) ; K )+ 1
S (t ^ (!) !)) :
;
e
Suppose 0
r(t )v (t ^
^
S (t ^ )) = IE e
r(t )v (t ^
^ ;
rT (S (T ) ; K )+ 1
S (T )<L
F (t) :
u t T . Then IE e
;
S (t ^ )) F (u)
f g
217
For 0
rt
Svx dB:
Integrate from 0 to t ^ :
r(t )v (t ^
^
+
Because e;r(t^ )v (t ^
Zt |0
ru
{z
Svx dB:
Zt
0
ru (;rv + v
2 2 t + rSvx + 1 S vxx ) du 2
is a martingale. Therefore,
S (t)vx(t S (t)) dB (t) 0 t Let X (t) be the wealth process corresponding to some portfolio (t). Then d(e rtX (t)) = e
We should take and Then
;
rt
rt
X (0) = v (0 S (0))
(t) = vx (t S (t)) 0 t T^ :
X (T ^ ) = v(T ^ S (T ^ )) ( v (T S (T )) = (S (T ) ; K )+ = v ( L) = 0
if if
>T T.
218
v(T, x)
v(t, x)
can become very negative near the knockout boundary. The hedger is in an unstable situation. He should take a large short position in the stock. If the stock does not cross the barrier L, he covers this short position with funds from the money market, pays off the option, and is left with zero. If the stock moves across the barrier, he is now in a region of (t) = v x(t S (t)) near zero. He should cover his short position with the money market. This is more expensive than before, because the stock price has risen, and consequently he is left with no money. However, the option has knocked out, so no money is needed to pay it off. Because a large short position is being taken, a small error in hedging can create a signicant effect. Here is a possible resolution. Rather than using the boundary condition
v(t L) = 0 0 t T
solve the PDE with the boundary condition
v(t L) + Lvx (t L) = 0 0 t T is a tolerance parameter, say 1%. At the boundary, Lvx (t L) is the dollar size of the
2. The value of the portfolio is always sufcient to cover a hedging error of size of the short position.
Chapter 21
Asian Options
Stock:
Payoff:
ZT
0
S (t) dt
X (0) = IE e
"
rT h
ZT
0
S (t) dt :
!#
S (t) = x Y (t) = y
we have the solutions
ZT
t
S (u) du:
220
v t S (t)
is the option value at time t, where The PDE for v is
Zt
0
;
S (u) du
;
v (t x y ) = e
r(T t) u(t
x y ):
(1.1)
v (T x y ) = h(y ) v(t 0 y) = e r(T t)h(y ): One can solve this equation rather than the equation for u.
; ;
dX = dS + r(X ; S ) dt = S (r dt + dB ) + rX dt ; r S dt = S dB + rX dt:
(t) is
We want to have
X (t) = v t S (t)
so that
Zt
0
S (u) du S (u) du
X (T ) = v T S (0)
=h
ZT
0
ZT
0
S (u) du :
221
dv t S (t)
Zt
0
S (u) du = vtdt + vxdS + vy S dt + 1 vxx dS dS 2 1 = (vt + rSvx + Svy + 2 2 S 2vxx ) dt + Svx dB = rv (t S (t)) dt + vx (t S (t)) S (t) dB (t): (From Eq. 1.1)
Take
dX (t) = rX (t) dt + (t) S (t) dB(t): (t) = vx (t S (t)): If X (0) = v (0 S (0) 0), then X (t) = v t S (t)
Zt
0
S (u) du
0 t T
because both these processes satisfy the same stochastic differential equation, starting from the same initial condition.
V =h
where 0 <
ZT
S (t) dt
; ;
< T . We compute v( x y ) = IE x y e r(T )h(Y (T )) just as before. For 0 t , we compute next the value of a derivative security which pays off v ( S ( ) 0)
at time . This value is The function w satises the Black-Scholes PDE
w(t x) = IE t xe
r( t) v (
;
S ( ) 0): x 0
r(T t)h(0)
;
0 t T: 0 t
< t T:
222 Remark 21.1 While no closed-form for the Asian option price is known, the Laplace transform (in 2 the variable 4 (T ; t)) has been computed. See H. Geman and M. Yor, Bessel processes, Asian options, and perpetuities, Math. Finance 3 (1993), 349375.
Chapter 22
Xk+1 =
k Sk+1 + (1 + r)(Xk ; k Sk ):
0
so that
1 (H )
X2(!1 !2 ) = V (!1 !2 ) 8! 1 !2 :
There are four unknowns: X0
(four equations)
1(H )
224
V (!1 H )
V (! 1 T )
The probabilities of the stock price paths are irrelevant, because we have a hedge which works on every path. From a practical point of view, what matters is that the paths in the model include all the possibilities. We want to nd a description of the paths in the model. They all have the property
n 1 X
;
The paths of log Sk accumulate quadratic variation at rate 2 per unit time.
k=0
If we change u, then we change , and the pricing and hedging formulas on the previous page will give different results. We reiterate that the probabilities are only introduced as an aid to understanding and computation. Recall:
Xk+1 =
k Sk+1 + (1 + r)(Xk ; k Sk ): k
Dene Then
= (1 + r)k :
Xk+1 =
k+1 k+1
Sk+1 + Xk ;
k+1 k k k+1
Sk
k
i.e.,
Xk+1 ; Xk =
k
Sk+1 ; Sk :
k
225
k f If we introduce a probability measure I under which Sk is a martingale, then P martingale, regardless of the portfolio used. Indeed,
Xk
k will also be a
k+1
=0
{z
}
Then we
X2 = V , where V
is some
X1 f X2 1 f V 1 + r X1 = 1 = IE 2 F 1 = IE 2 F 1 f f X0 = X0 = IE X1 = IE V :
0 1 2
To nd the risk-neutral probability measure
f IP f!k = H g, q = IP f!k = T g, and compute ~ f f ~ ~ IE Sk+1 F k = pu Sk + qd Sk k+1 k+1 k+1 1 pu + q d] Sk : = 1+r ~ ~ k We need to choose p and q so that ~ ~
pu + qd = 1 + r ~ ~ p + q = 1: ~ ~
The solution of these equations is
is a martingale, we denote
p= ~
r p = 1 + ;; d ~ u d
q = u ; (1 + r) : ~ u;d
226 Let
B (t) 0 t T , be a Brownian motion dened on a probability space ( F P). 2 IR, the paths of t + B(t) S (t) = S (0) expf t + B(t)g
For any
accumulate quadratic variation at rate 2 per unit time. Surprisingly, the choice of in this denition is irrelevant. Roughly, the reason for this is the following: Choose ! 1 2 . Then, for 1 2 IR,
1t +
1t +
B (t !1 ) = 2t + B (t !2) 0 t T:
In other words, regardless of whether we use 1 or 2 in the denition of S (t), we will see the same paths. The mathematically precise statement is the following: If a set of stock price paths has a positive probability when S (t) is dened by
1 = r; 2
S (t) = S (0) expfrt + B(t) ; 1 2tg 2 e rt S (t) = S (0) expf B(t) ; 1 2tg 2
;
227
h |
r dt + dB (t)
f We can change to the risk-neutral measure I , under which P Brownian motion, and then proceed as if had been chosen to be equal to r ; 1 2 . 2 e B has the same paths as B .
22.3 Risk-neutral pricing and hedging
{z e dB (t)
} e B is a
e dS (t) = rS (t) dt + S (t) dB (t) e f P where B is a Brownian motion under I . Set (t) = ert :
Then
(3.1)
(3.2)
f Regardless of the portfolio used, X((tt)) is a martingale under I . P Now suppose V is a given F (T )-measurable random variable, the payoff of a simple European derivative security. We want to nd the portfolio process (T ) 0 t T , and initial portfolio X (t) must be a martingale, we must have value X (0) so that X (T ) = V . Because (t)
X (t) = IE V F (t) f (t) (T )
0 t T:
(3.3) This is the risk-neutral pricing formula. We have the following sequence:
228 1.
2. Dene X (t)
is given,
0 t T , by (3.3) (not by (3.1) or (3.2), because we do not yet have (t)). (t) so that (3.2) (or equivalently, (3.1)) is satised by the X (t) 0 t T ,
To carry out step 3, we rst use the tower property to show that X((tt)) dened by (3.3) is a martingale f P under I . We next use the corollary to the Martingale Representation Theorem (Homework Problem 4.5) to show that
(3.4)
for some proecss . Comparing (3.4), which we know, and (3.2), which we want, we decide to dene (3.5)
T , is the value of
f V X (0) = IE (T )
V f V X (T ) = (T )IE (T ) F (T ) = (T ) (T ) = V: Remark 22.1 Although we have taken r and to be constant, the risk-neutral pricing formula is still valid when r and are processes adapted to the ltration generated by B . If they depend on e either B or on S , they are adapted to the ltration generated by B . The validity of the risk-neutral
pricing formula means: 1. If you start with
229
To see if the risk-neutral measure is unique, compute the differential of all discounted asset prices e e and check if there is more than one way to dene B so that all these differentials have only d B terms.
= h(S(T)). We can take the stock price to be the state i e h v(t x) = IE t x e;r(T ;t) h(S(T)) :
Then
V =h
Take S(t) and Y (t) = 0
ZT
0
S(u) du :
h(Y (T))
where
ZT
t
S(u) du:
230
Then
e dS(t) = r(t Y (t)) S(t)dt + (t Y (t))S(t) dB(t) e dY (t) = (t Y (t)) dt + (t Y (t)) dB(t) V = h(S(T)):
Then
is a martingale under IP .
In every case, we get an expression involving v to be a martingale. We take the differential and set the dt term to zero. This gives us a partial differential equation for v , and this equation must e hold wherever the state processes can be. The dB term in the differential of the equation is the differential of a martingale, and since the martingale is
231
1 d X(t) = (t) ;r(t Y (t))v(t S(t) Y (t)) dt (t) + vt dt + vx dS + vy dY + 1 vxx dS dS + vxy dS dY + 1 vyy dY dY 2 2 1 1 = (t) (;rv + vt + rSvx + vy + 1 2 S 2 vxx + Svxy + 2 2 vyy ) dt 2 e + ( Svx + vy ) dB
The partial differential equation satised by v is
xvxy + 1 2 vyy = 0 2
= v(t x y), and all other variables are functions of (t y). We have 1 e d X(t) = (t) Svx + vy ] dB(t) (t) = (t Y (t)), = (t Y (t)), v = v(t S(t) Y (t)), and S = S(t). We want to choose (t) so that (t) to be
232
Chapter 23
be a process on
Then B is a Brownian motion. Proof: (Idea) Let 0 s < t T be given. We need to show that B (t) ; B (s) is normal, with mean zero and variance t ; s, and B (t) ; B (s) is independent of F (s). We shall show that the conditional moment generating function of B (t) ; B (s) is
IE eu(B (t)
B (s)) F (s)
1 2 = e 2 u (t s) :
;
Since the moment generating function characterizes the distribution, this shows that B (t) ; B (s) is normal with mean 0 and variance t ; s, and conditioning on F (s) does not affect this, i.e., B (t) ; B(s) is independent of F (s). We compute (this uses the continuity condition (1) of the theorem)
Zt
s
Zt
s
euB(v)
uses cond. 3
dv: |{z}
234
Zt
s
=; = 0:
It follows that
Zs
0
Zt
0
ueuB(v)dB (v ) F (s)
Zt
s
'(s) = euB(s)
and
Zt
s
'(v) dv
Plugging in s, we get
1 u2 s 2 :
s)
B (s)) F (s)
1 2 = e 2 u ( t s) :
;
235
dS1 = r dt + S1 dS2 = r dt + S
2
Dene
2 + 2 1= q 11 12 2 + 2 2= 21 22 11 21 + 12 =
Dene processes W1 and W2 by
1 2
22 :
dW1 = dW2 =
12 dB2 )2 2 dB dB ) 12 2 2
dB1 +
and similarly Therefore, W1 and W2 are Brownian motions. The stock prices have the representation
dS1 = r dt + S1 dS2 = r dt + S
2
11 21 + 12 22 ) dt
236
dS1 = r dt + dW 1 1 S1 dS2 = r dt + dW 2 2 S
2
=
so that
0
"
11 21 11 12
12 22 21 22
= = =
11 21
2 + 2 11 12 + 12 22 11 21 # 2 1 2 1 2 1 2 2 11 = 1
12 22
#"
# #
11 21 + 12 22 2 + 2 21 22
12 = 0 22 =
21 =
This corresponds to
1;
2 2:
p 2 =) dB2 = dW2 ; dW1 1; If = 1, then there is no B2 and dW2 = dB1 = dW1: Continuing in the case 6= 1, we have
dB2
( 6= 1)
dB1 dB1 = dW1 dW1 = dt 1 dB2 dB2 = 1 ; 2 dW2 dW2 ; 2 dW1 dW2 + 2dW2 dW2
1 = 1; = dt
2
dt ; 2 2 dt + 2 dt
237
dB1 dB2 = p 1
1; =p 1 1;
2 (dW1 dW2 ; 2(
dW1 dW1 )
dt ; dt) = 0:
We can now apply an Extension of Levys Theorem that says that Brownian motions with zero cross-variation are independent, to conclude that B 1 B2 are independent Brownians.
238
Chapter 24
dY (t) = dt + Y (t)
Stock process:
1 dB1 (t):
dS (t) = dt + S (t)
where 1
2 dB1 (t) + 1 ;
2 2 dB2 (t)
> 0 2 > 0 ;1 < < 1, and B1 and B2 are independent Brownian motions on some ( F P). The option pays off: (S (T ) ; K )+ 1 Y (T )<L
f g
at time T , where
Y (T ) = 0maxT Y (t): t
Remark 24.1 The option payoff depends on both the Y and S processes. In order to hedge it, we will need the money market and two other assets, which we take to be Y and S . The risk-neutral measure must make the discounted value of every traded asset be a martingale, which in this case means the discounted Y and S processes. We want to nd 1 and 2 and dene
e dB1 = 1 dt + dB1
e dB2 =
239
2 dt + dB2
240 so that
dY = r dt + dB 1 e1 Y = r dt + 1 1 dt + 1 dB1 dS = r dt + dB + q1 ; 2 e1 S
2 2 dB2 e
We must have
(0.1)
1;
2 2 2:
(0.2)
We solve to get
;r; 2= p 1; 2 n
= ;r
1
2 1: 2
We shall see that the formulas for 1 and 2 do not matter. What matters is that (0.1) and (0.2) uniquely determine 1 and 2 . This implies the existence and uniqueness of the risk-neutral measure. We dene
f e e P Under I , B1 and B2 are independent Brownian motions (Girsanovs Theorem). risk-neutral measure. f P Remark 24.2 Under both IP and I , Y has volatility 1, S has volatility 2 and
i.e., the correlation between dY and dS is . Y S The value of the option at time zero is
f IP is the unique
dY dS = YS
1 2 dt
rT (S (T ) ; K )+ 1
Y (T )<L
i
g
We need to work out a density which permits us to compute the right-hand side.
241
dY = r dt + Y
so
e 1 dB1 n o
b c B (T ) and M (T ), appearing in Chapter 20, is f b IP fB (T ) 2 d^ M (T ) 2 dmg b c ^ ^) ( (2m ; ^)2 b^ 1 b2 ) ^ m ^ = 2(2p ; b exp ; ^2T b + b ; 2 T db dm ^ T 2 T m > 0 ^ < m: ^ b ^ q e1 + 1 ; 2 dB
2 2 dB2 e
dS = r dt + S
so
e b B1 (T ) = ; bT + B (T )
so
b S (T ) = S (0) expfrT + 2B (T ) ; 1 2 T ; 2 2
q bT + 1 ; 2
2 2B2 (T )g e
242
f v (0 S (0) Y (0)) = IE e
;
rT (S (T ) ; K )+ 1
Y (T )<L
i
g
1 f = e rT IE S (0) exp (r ; 2 2 ; 2
2 b)T +
b 2 B (T ) + 1 ;
2 2 B2 (T ) e
;K
:1 Y (0)exp 1 M (T )]<L b
f
( ) b b b ffB2(T ) 2 d~g = p 1 exp ; ~2 d~ ~ 2 IR: e IP b 2T 2 T e e b c Furthermore, the pair of random variables (B (T ) M (T )) is independent of B2 (T ) because B1 and e2 are independent under IP . Therefore, the joint density of the random vector (B2(T ) B (T ) M (T )) f e b c B
is
v (0 S (0) Y (0))
=e
;
rT
1 1
log Y L Zm Z (0) ^ 0
S (0) exp (r ;
1 2; 2 2
2 b)T +
b 2^ + 1 ;
2 2~ b
;K
T 2 T :d~ d^ dm: b b ^
The answer depends on T S (0) and Y (0). It also depends on 1 2 r K and L. It does not depend on 1 nor 2 . The parameter b appearing in the answer is b = r1 ; 21 : Remark 24.3 If we had not regarded Y as a traded asset, then we would not have tried to set its mean return equal to r. We would have had only one equation (see Eqs (0.1),(0.2))
= r+
1+ 1;
2 2 2
(1.1)
to determine 1 and 2 . The nonuniqueness of the solution alerts us that some options cannot be hedged. Indeed, any option whose payoff depends on Y cannot be hedged when we are allowed to trade only in the stock.
243
S , then Y
1;
dS = dt + S
we should set
2 dB1 +
2 2 dB2
= ;r
f so with dW
= dt + dW
we have
dS = r dt + S
and we are on our way.
f dW
(S (T ) ; K )+ 1 Y (T )<L
f
v (t x y ) = e
r(T t)I t x y f E
;
(S (T ) ; K )+ 1 maxt
f
u T Y (u) < Lg
by replacing T , S (0) and Y (0) by T ; t, x and y respectively in the formula for v (0 S (0) Y (0)). Now start at time 0 at S (0) and Y (0). Using the Markov property, we can show that the stochastic process
e rt v (t S (t) Y (t))
;
e 2 Svx dB1 +
2 2 1 2 SY vxy + 1 1 Y vyy 2
dt
1;
244
y L v(t, x, L) = 0, x >= 0
x v(t, 0, 0) = 0
Figure 24.1: Boundary conditions for barrier option. Note that t 2
0 T ] is xed.
2 2 1 2 xyvxy + 1 1 y vyy 2
=0 0 t<T
x 0 0 y L:
v(T x y) = (x ; K )+
x 0 0 y<L
v(t 0 0) = 0 0 t T v (t x L) = 0 0 t T x 0:
245
x=0
v (t 0 L) = 0 v(t 0 0) = 0 v (T 0 y) = (0 ; K )+ = 0 y 0:
On the x = 0 boundary, the option value is v (t 0 y ) = 0 0 y L:
v(t 0 0) = e
r(T t)(0 ; K )+ = 0
;
v (T x 0) = (x ; K )+
x 0:
On the y = 0 boundary, the barrier is irrelevant, and the option value is given by the usual Black-Scholes formula for a European call.
= e rt
;
e 2Svx dB1 + 1 ; i i
where vx that
rt
d e rt Y (t) = e =e
;
= e rt
; ;
;rS (t) dt + dS (t)] q e 2S (t) dB1 (t) + 1 ; ;rY (t) dt + dY (t)] e 1Y (t) dB1 (t):
; ;
rt
rt
Therefore,
Let 2 (t) denote the number of shares of stock held at time t, and let 1(t) denote the number of shares of the barrier process Y . The value X (t) of the portfolio has the differential
dX =
2dS + 1 dY
+rX;
2 S ; 1Y ] dt:
d e rtX (t)] =
;
2 (t)d e
rtS (t)] +
1(t)d e
rt Y (t)]:
To get X (t) = v (t
and
2 (t) = vx (t
Chapter 25
American Options
This and the following chapters form part of the course Stochastic Differential Equations for Finance II.
dS = rS dt + S dB
Intrinsic value at time t : (K ; S (t))+ : Let L 2
0 K ] be given. Suppose we exercise the rst time the stock price is L or lower. We dene
vL
K;x (K ; L)IEe
L ))
Lg
+
if x L, if x > L:
rL
The plan is to comute vL (x) and then maximize over L to nd the optimal exercise price. We need to know the distribution of L .
248
Intrinsic value
Stock price
m > 0 b < m:
Z Z m 2(2m ; b) ( (2m ; b)2 ) p IP fM (t) xg = db dm exp ; 2t t 2 t x ) ( Z 2 exp ; (2m ; b)2 b=m dm p = 2t 2 t x b= ( 2) Z p 2 exp ; mt dm: = 2 2 t
1 ;1 1 ;1 1
2 p2 exp ; z2 dz: = x= t 2
1 p
Now
t()M (t) x
249
IEe
=e x
;
e t IP f 2 dtg
;
> 0:
(See Homework)
Reference: Karatzas and Shreve, Brownian Motion and Stochastic Calculus, pp 95-96.
t < 1, dene
e B (t) = t + B (t) Z (t) = expf; B (t) ; 1 2 tg 2 e (t) + 1 2tg = expf; B 2 e ~ = minft 0 B (t) = xg:
A
Dene
We x a nite time T and change the probability measure only up to T . More specically, with T xed, dene Z
f IP (A) =
Z (T ) dP
A 2 F (T ):
T we have
IP f~ tg = IE 1
h h
f = IE 1
~ t
i
g g
f = IE 1
i fh e = IE 1 ~ t expf B (~ ^ t) ; 1 2 (~ ^ t)g 2 i fh = IE 1 ~ t expf x ; 1 2 ~g 2 Zt 1 f = expf x ; 2 2 sgIP f ~ 2 dsg 0 ( Zt x 2) 1 2s ; x p exp x ; 2 = 2s ds 0 s 2 s ) ( Zt x (x ; s)2 ds: p exp ; 2s = s 2 s
f g f g f g
1 2 e ~ t expf B (T ) ; 2 T g 1 f f e = IE 1 ~ t IE expf B (T ) ; 2 2 T g F (~ ^ t)
f g
~ t
Z (T )
Therefore,
= minft 0 B (t) = xg
for nondrifted Brownian motion:
IEe
For
Z
0
2 px exp ; t ; xt dt = e 2 t 2 t
x 2
p
> 0 x > 0:
~ = minft 0 t + B (t) = xg
251
IEe
~=
where in the last step we have used the formula for IEe; If
with
replaced by
+1 2
2.
lim e
;
if
~(! ) = 1, then e
#0 #0
~(! ) = 1
Therefore,
lim e
Letting
#0 and using the Monotone Convergence Theorem in the Laplace transform formula
IEe
;
#0
~(! ) = 1
~<
:
2
~ = ex x 2 +
p ; ;
we obtain If If
IP f~ < 1g = ex
= ex x :
; j j
0, then
< 0, then
252 is a martingale, so Y (t ^
expf B (t ^ ) ; 1 2 (t ^ )g 2
1
ex
1 = IE t! expf B (t ^ ) ; 1 2 (t ^ )g: lim 2 There are two possibilities. For those ! for which (! ) < 1,
; 1
1 2 lim expf B (t ^ ) ; 1 2 (t ^ )g = e x 2 : 2 t! For those ! for which (! ) = 1, 1 lim expf B (t ^ ) ; 1 2 (t ^ )g t! expf x ; 2 2tg = 0: lim 2 t!
1 1
Therefore,
= IE e x
;
1 2
1<
1 where we understand e x; 2
Let
= IEe x
2
1 2
=1 2
2 , so
= 2
to be zero if
= 1.
;
x 2
p
= IEe
>0 x>0
253
(K ; S (t)) +.
Dene
p ; ; 1 2r +
2 = ; r + 1 ; 1 2r + r ; =2 2 2
s s s s
= ; r2 + 1 ; 1 2r + r 2 ; r + 2 =4 2
2 2 = ; r2 + 1 ; 1 r 2 + r + 2=4 2
= ; r2 + 1 ; 1 r + =2 2 = ; r2 + 1 ; 1 r + =2 2 = ; 2r :
2
Therefore,
8 <(K ; x) vL (x) = : ;x (K ; L) L
;
2r=
x L:
2
0 x L
;x The curves (K ; L) L
2r=
C = (K ; L)L2r=
254
value
K-x K 2 (K - L) (x/L)-2r/
Stock price
value
C3 x C2 x C1 x
-2r/ 2
-2r/ 2 -2r/ 2
Stock price
255
We solve
; 1 + 2r + 2r K = 0 2 2L
to get
2 L = 2 rK2r : +
Since 0 < 2r < 2 + 2r we have
0 < L < K: 0 x L
Solution to the perpetual American put pricing problem (see Fig. 25.4):
2r=
x L
where
2 L = 2 rK2r : +
Note that
v (x) = ;12r
0
; 2 (K ; L) (L
0
)2r=
2r=
2;
0 x<L
x>L :
We have
x#L
lim v (x) = ;2 r2 (K ; L ) 1
L 2+ 2 = ;2 r2 K ; 2 rK2r 2rK2r + 2 + 2r ; 2r ! 2 + 2r = ;2 r2 2 + 2r 2r
= ;1 = lim v (x):
x"L
0
256
value
L*
Stock price
= 2r 2
(7.1) (7.2)
= IE x
where
e (K ; L )+1
;
C r
<
1g
S (0) = x
If 0
= minft 0 S (t) = L g:
00
(7.3) (7.4)
x < L , then ;rv(x) + rxv (x) + 1 2x2v (x) = ;r(K ; x) + rx(;1) = ;rK: 2 If L x < 1, then ;rv(x) + rxv (x) + 1 2x2v (x) 2 = C ;rx ; rx x 1 ; 1 2 x2 (; ; 1)x 2 ] 2 1 2 (; ; 1)] = Cx ;r ; r ; 2 = C (; ; 1)x r ; 1 2 2r 2 2 = 0:
0 0 00 ; ; ; ; ; ; ;
In other words, v solves the linear complementarity problem: (See Fig. 25.5).
257
6
K@
@@
L
@@ @@
-x
Figure 25.5: Linear complementarity
and L is the boundary between them. If the stock price is in C , the owner of the put should not exercise (should continue). If the stock price is in S or at L , the owner of the put should exercise (should stop).
rt
d e rtv(S (t)) = e
;
rt
We should set
As long as the owner does not exercise, you can consume the interest from the money market position, i.e.,
v(S (t)) = K ; S (t) (t) = v (S (t)) = ;1: To hedge the put when S (t) < L , short one share of stock and hold K in the money market.
0
C (t) = rK 1 S(t)<L :
f g
e rtv (S (t)) is a supermartingale (see its differential above). e rtv (S (t)) e rt (K ; S (t))+ , 0 t < 1; e rtv (S (t)) is the smallest process with properties 1 and 2.
; ; ; ;
Y (t) e rt (K ; S (t))+
;
0 t < 1:
(8.1)
Y (t) e rt v (S (t)) 0 t < 1: (8.2) We use (8.1) to prove (8.2) for t = 0, i.e., Y (0) v(S (0)): (8.3) If t is not zero, we can take t to be the initial time and S (t) to be the initial stock price, and then
;
adapt the argument below to prove property (8.2). Proof of (8.3), assuming Y is a supermartingale satisfying (8.1): Case I: S (0)
L : We have
259
> 0, we have
f 1g
Y (0)
2 IE 6e r (K ; S{z ))+ 1 4 |( }
f 1g ;
= v (S (0)):
<
1g
3 7 5
(by 8.1)
where the supremum is over all stopping times. Optimal exercise rule: Any stopping time Characterization of v : 1. 2. 3. which attains the supremum.
e rtv (S (t)) is a supermartingale; e rtv (S (t)) e rt h(S (t)) 0 < t < 1; e rtv (S (t)) is the smallest process with properties 1 and 2.
; ; ; ;
v(x) = sup IE x e r (S ( ) ; K )+
;
Theorem 10.63
v (x) = x 8x 0:
Then: 1. 2.
Therefore, Y (0)
x v(x):
we choose,
;
IE x e r (S ( ) ; K )+ < IE x e r S ( )
There is no optimal exercise time.
x = v(x):
:stopping time
See Fig. 25.6. It can be shown that jump. Let S (0) be given. Then
| {z }
261
v (T x) = 0 x K
v (T x) = K ; x 0 x K
-t
e rtv (t S (t)) 0 t T is a supermartingale; e rtv (t S (t)) e rt (K ; S (t))+ 0 t T ; e rtv (t S (t)) is the smallest process with properties 1 and 2.
; ; ;
v(t x) = sup IE x e
t T
Then
r( t)h(S (
;
)):
At every point (t
x) 2 0 T ]
262 1. 2. 3.
e rtv (t S (t)) 0 t T is a supermartingale; e rtv (t S (t)) e rt h(S (t)); e rtv (t S (t)) is the smallest process with properties 1 and 2.
; ; ; ;
(!) = 1, then there is no optimal exercise time along the particular path ! .
Chapter 26
dS (t) = r(t)S (t) dt + (t)S (t) dB (t) where r and are processes and r(t) 0 0 t T a.s. This stock pays no dividends. 0, and assume h(0) = 0. (E.g., h(x) = (x ; K )+ ). An Let h(x) be a convex function of x American contingent claim paying h(S (t)) if exercised at time t does not need to be exercised
before expiration, i.e., waiting until expiration to decide whether to exercise entails no loss of value. Proof: For 0
1 and x 0, we have
h( x) = h((1 ; )0 + x) (1 ; )h(0) + h(x) = h(x): Let T be the time of expiration of the contingent claim. For 0 t T ,
0
and S (T )
0, so
(*)
Consider a European contingent claim paying h(S (T )) at time T . The value of this claim at time t 2 0 T ] is
264
.... .... .... .. ......r.. . ..... . ..... ....... (x h(x)) .. h(x) ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ...... ...... ...... ...... ....... ....... ....... ............r....... .. ............ ...... . . . . . .... . . .. . . . . h( x) ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... ...... ...... ...... ...... ................................. ...................r.......... .. h . . . . . . . . ....... . . . . ........ . . ......... . . . . . . ........... . . ............ . . ..............
X (t) = 1 IE (t) h(S (T )) F (t) (t) (t) (T ) (t) 1 (t) IE h (T ) S (T ) F (t) (by (*)) 1 h (t) IE S (T ) F (t) (Jensens inequality) (t) (T ) = 1 h (t) S (t) ( S is a martingale) (t) (t)
1 = (t) h(S (t)):
This shows that the value X (t) of the European contingent claim dominates the intrinsic value h(S (t)) of the American claim. In fact, except in degenerate cases, the inequality
0 < < 1:
265
1 2 0 t t1 S (t) = S (0) expf(r ; 2 )t +1 B(t)g 2 )(t ; t1 ) + (B (t) ; B (t1 ))g (1 ; )S (t1) expf(r ; 2 t1 < t T: Consider an American call on this stock. At times t 2 (t 1 T ), it is not optimal to exercise, so the
value of the call is given by the usual Black-Scholes formula
r(T t) N (d
;
(T ; t x))
t1 < t T :
At time t1 , immediately after payment of the dividend, the value of the call is
2=2)
At time t1 , immediately before payment of the dividend, the value of the call is
w(t1 S (t1))
where Theorem 2.65 For 0
w(t1 x) = max (x ; K )+ v (t1 (1 ; )x : t t1 , the value of the American call is w(t S (t)), where w(t x) = IE t x e
r(t1 t) w(t
;
S (t1)) :
0 t t1
t1 < t T:
Proof: We only need to show that an American contingent claim with payoff w(t 1 S (t1)) at time t1 need not be exercised before time t1. According to Theorem 1.64, it sufces to prove 1.
w(t1 0) = 0,
266 2.
w(t1 x) is convex in x.
0) = 0, we have immediately that
Since v (t1
is the number of shares of stock held by the hedge immediately after payment of the dividend. The post-dividend position can be achieved by reinvesting in stock the dividends received on the stock held in the hedge. Indeed,
(t1)
dividends received price per share when dividend is reinvested
Case II: v (t1 (1 ; )x) < (x ; K )+ . The owner of the option should exercise before the dividend payment at time t 1 and receive (x ; K ). The hedge has been constructed so the seller of the option has x ; K before the dividend payment at time t1 . If the option is not exercised, its value drops from x ; K to v (t 1 (1 ; )x), and the seller of the option can pocket the difference and continue the hedge.
Chapter 27
F P). Con-
(t) = exp
Zt
0
r(u) du :
A zero-coupon bond, maturing at time T , pays 1 at time T and nothing before time T . According to the risk-neutral pricing formula, its value at time t 2 0 T ] is
ZT
t
r(u) du F (t) :
Given B (t
T ) dollars at time t, one can construct a portfolio of investment in the stock and money
267
268 market so that the portfolio value at time T is 1 almost surely. Indeed, for some process ,
dB (t T ) = r(t) (t) B (0 T ) + (u) dW (u) dt + (t) (t) dW (t) 0 = r(t)B (t T ) dt + (t) (t) dW (t):
The value of a portfolio satises
Zt
dX (t) = (t) dS (t) + r(t) X (t) ; (t)S (t)]dt = r(t)X (t) dt + (t) (t)S (t) dW (t):
(*) We set
( ZT ) B (t T ) = exp ; r(u) du
t
dB (t T ) = r(t)B(t T ) dt
i.e., = 0. Then given above is zero. If, at time t, you are given B (t invest only in the money market, then at time T you will have
B(t T ) exp
(Z T
t
r(u) du = 1:
If r(t) is random for all t, then is not zero. One generally has three different instruments: the stock, the money market, and the zero coupon bond. Any two of them are sufcient for hedging, and the two which are most convenient can depend on the instrument being hedged.
269
IE (1 ) (S (T ) ; F (t)) F (t) = 0 0 t T: T
0 = IE 1 (S (T ) ; F (t)) F (t) (T ) = IE S (T ) F (t) ; F (t) IE (t) F (t) (T ) (t) (T ) = S (t) ; F (t) B (t T ): (t) (t)
) F (t) = BSt(tT ) : (
Remark 27.1 (Value vs. Forward price) The T -forward price F (t) is not the value at time t of the forward contract. The value of the contract at time t is zero. F (t) is the price agreed upon at time t which will be paid for the stock at time T .
V (t) = (t) IE (1 ) (S (T ) ; F (0)) F (t) T t (T ) = (t) IE S(T ) F (t) ; F (0) IE ((T)) F (t) (t) = (t) S(t) ; F (0)B (t T ) = S (t) ; F (0)B (t T ): F (0)B (0 T ) = BS (0) ) B (0 T ) = S (0): (0 T
This suggests the following hedge of a short position in the forward contract. At time 0, short F (0) T -maturity zero-coupon bonds. This generates income
270 Buy one share of stock. This portfolio requires no initial investment. Maintain this position until time T , when the portfolio is worth
S (T ) ; F (0)B (T T ) = S (T ) ; F (0):
Deliver the share of stock and receive payment F (0). A short position in the forward could also be hedged using the stock and money market, but the implementation of this hedge would require a term-structure model.
(tk+1 ) = exp
Z tk+1
r(u) du
Enter a future contract at time tk , taking the long position, when the future price is (t k ). At time tk+1 , when the future price is (tk+1 ), you receive a payment (tk+1 ) ; (tk ). (If the price has fallen, you make the payment ;( (tk+1 ) ; (tk )). ) The mechanism for receiving and making these payments is the margin account held by the broker. By time T
is F (tk )-measurable.
= tn , you have received the sequence of payments (tk+1 ) ; (tk ) (tk+2 ) ; (tk+1 ) : : : (tn ) ; (tn 1 )
;
2n 1 X (t) IE 4
;
j =k
Because it costs nothing to enter the future contract at time t, this expression must be zero almost surely.
271
(t) IE
"Z T
t
0 t T:
Note that (tj +1 ) appearing in the discrete-time version is F (t j )-measurable, as it should be when approximating a stochastic integral. Denition 27.1 The T -future price of the stock is any F (t)-adapted stochastic process
f (t) 0 t T g
satisfying
IE
"Z T
t
(T ) = S (T ) a.s., and
(a) (b)
0 t T:
(t) = IE S (T ) F (t) 0 t T:
Proof: We rst show that (b) holds if and only if R t 1 d (u) is also a martingale, so 0 (u) is a martingale. If is a martingale, then
IE
"Z T
t
"Z T
0
satises
(u)
d (u) 0 t T:
this implies
Now dene
(t) = IE S (T ) F (t)
0 t T:
Clearly (a) is satised. By the tower property, is the only martingale satisfying (a).
ZT
0
Thus, if the future contract holder takes delivery at time T , he has paid a total of for an asset valued at S (T ).
(t) = IE S (T ) F (t)
) F (t) = BSt(tT ) = (
(t)IE
S (t)
1 (T )
F (t)
Forward-future spread:
IE
IE
(T )
1 (T )
IE
1 IE (S (T )) ; IE S (T ) (T ) (T )
(0) = F (0):
273
(0) F (0):
This is the case that a rise in stock price tends to occur with a fall in the interest rate. The owner of the future tends to receive income when the stock price rises, but invests it at a declining interest rate. If the stock price falls, the owner usually must make payments on the future contract. He withdraws from the money market to do this just as the interest rate rises. In short, the long position in the future is hurt by positive correlation between (1 ) and S (T ). The buyer of the future is T compensated by a reduction of the future price below the forward price.
Because (1 ) T
(t) = ert 1 S (t) = S (0) expf( ; 2 2)t + W (t)g f = S (0) expf(r ; 1 2 )t + W (t)g 2
f (t) = IE S (T ) F (t)]
= F (t) ) = BSt(tT ) = er(T t)S (t): (
;
oi
If > r, then (0) < IES (T ): This situation is called normal backwardation (see Hull). If then (0) > IES (T ). This is called contango.
< r,
Chapter 28
Term-structure models
Throughout this discussion, fW (t) 0 t T g is a Brownian motion on some probability space ( F P), and fF (t) 0 t T g is the ltration generated by W .
Suppose we are given an adapted interest rate process fr(t) lation factor Z
(t) = exp
r(u) du
0 t T:
In a term-structure model, we take the zero-coupon bonds (zeroes) of various maturities to be the primitive assets. We assume these bonds are default-free and pay $1 at maturity. For 0 t T T , let B(t T ) = price at time t of the zero-coupon bond paying $1 at time T . Theorem 0.67 (Fundamental Theorem of Asset Pricing) A term structure model is free of arbif trage if and only if there is a probability measure I on (a risk-neutral measure) with the same P probability-zero sets as IP (i.e., equivalent to IP ), such that for each T 2 (0 T ], the process
B(t T ) (t)
0 t T
f is a martingale under I . P
Remark 28.1 We shall always have
dB (t T ) = (t T )B (t T ) dt + (t T )B(t T ) dW (t) 0 t T
for some functions
(t T ) and (t T ). Therefore
276 so IP is a risk-neutral measure if and only if (t T ), the mean rate of return of B (t T ) under IP , is the interest rate r(t). If the mean rate of return of B (t T ) under IP is not r(t) at each time t and for f each maturity T , we should change to a measure I under which the mean rate of return is r(t). If P such a measure does not exist, then the model admits an arbitrage by trading in zero-coupon bonds.
fr(t) 0 t )
ZT
t
r(u) du F (t)
0 t T
T:
k:t<Tk
Pk B (t Tk ):
g
T.
< T.
(t) IE
1 (B (T T ) ; K )+ F (t) 1 (T1)
0 t T1:
277
28.3 Terminology
Denition 28.1 (Term-structure model) Any mathematical model which determines, at least theoretically, the stochastic processes
B(t T ) 0 t T
for all T
2 (0 T ].
t T T
, the yield to maturity
Y (t T ) is the
Y (t T ) = ; T 1 t log B(t T ): ; B (t T ) 0 t T Y (t T ) 0 t T T T:
B (t T ) = expf R(t T T + )g B (t T + )
or equivalently,
#0
(4.1)
278 This is the instantaneous interest rate, agreed upon at time t, for money borrowed at time T . Integrating the above equation, we obtain
ZT
t
f (t u) du = ;
ZT @ @u log B(t u) du
t u=T u=t
= ; log B (t u) = ; log B (t T )
so
( ZT ) B(t T ) = exp ; f (t u) du :
t
You can agree at time t to receive interest rate f (t u) at each time u 2 t T ]. If you invest $ B (t at time t and receive interest rate f (t u) at each time u between t and T , this will grow to
T)
B(t T ) exp
at time T .
(Z T
t
f (t u) du = 1
"
( ZT ) B (t T ) = exp ; f (t u) du t ( ) @ B (t T ) = ;f (t T ) exp ; Z T f (t u) du @T
t
279
(u T ) du +
Here f
(u T ) dW (u)
0 t T:
df (t T ) = (t T ) dt + (t T ) dW (t):
( ZT ) B(t T ) = exp ; f (t u) du :
t
Now
Let Then
B (t T ) = g ;
and
ZT
t
f (t u) du
! !
dt ; dW )
dB (t T ) = dg ;
=g ;
0
ZT
t
00
t ZT
f (t u) du
f (t u) du (r dt ;
1g 2
ZT
t
f (t u) du ( )2 dt
280
0 t T
T T: T:
(7.1)
ZT
t
(t u) du =
1 2
ZT
t
(t u) du
!2
0 t T 0 t T
T , we obtain
(t T ) = (t T )
ZT
t
(t u) du
(7.2)
Not only does (7.1) imply (7.2), (7.2) also implies (7.1). This will be a homework problem. Suppose (7.1) does not hold. Then IP is not a risk-neutral measure, but there might still be a riskneutral measure. Let f (t) 0 t T g be an adapted process, and dene
f W (t) = f IP (A) =
Then
Zt
0
(u) du + W (t)
Z (t) = exp ;
Zt
0
(u) dW (u) ;
Z (T ) dIP 8A 2 F (T ):
Zt 2 1 2 0 (u) du i
dB (t T ) = B(t T ) r(t) ; (t T ) + 1 ( (t T ))2 dt 2 ; (h T )B(t T ) dW (t) t i = B (t T ) r(t) ; (t T ) + 1 ( (t T ))2 + (t T ) (t) dt 2 f ; (t T )B(t T ) dW (t) 0 t T: f In order for B (t T ) to have mean rate of return r(t) under I , we must have P (t T ) = 1 ( (t T ))2 + (t T ) (t) 0 t T T : 2 Differentiation w.r.t. T yields the equivalent condition (t T ) = (t T ) (t T ) + (t T ) (t) 0 t T T :
Theorem 7.68 (Heath-Jarrow-Morton) For each T 2 (0 (u T ) 0 u T , be adapted processes, and assume f (0 T ) 0 t T , be a deterministic function, and dene
(7.3)
f (t T ) = f (0 T ) +
Zt
0
(u T ) du +
Zt
0
281
t T T is a family of forward rate processes for a term-structure model Then f (t T ) 0 without arbitrage if and only if there is an adapted process (t) 0 t T , satisfying (7.3), or equivalently, satisfying (7.4).
Remark 28.2 Under IP , the zero-coupon bond with maturity T has mean rate of return
r(t) ; (t T ) + 1 ( (t T ))2 2
and volatility
and when normalized by the volatility, this becomes the market price of risk
1 ; (t T ) + 2 ( (t T ))2 :
(t T )
The no-arbitrage condition is that this market price of risk at time t does not depend on the maturity T of the bond. We can then set
; (t T ) + 1 ( (t T ))2 2 (t) = ; (t T )
and (7.3) is satised. (The remainder of this chapter was taught Mar 21)
"
Suppose the market price of risk does not depend on the maturity T , so we can solve (7.3) for . Plugging this into the stochastic differential equation for B (t T ), we obtain for every maturity T :
(t T ) 0 t T T (t) 0 t T :
282 These may be stochastic processes, but are usually taken to be deterministic functions. Dene
(t T ) = (t T ) (t T ) + (t T ) (t)
f W (t) =
Zt
0
(u) du + W (t)
Z f(A) = Z (T ) dIP 8A 2 F (T ): IP
0
Let f (0
Z (t) = exp ;
A
Zt
(u) dW (u) ; 1 2
Zt
0
2 (u) du
T) 0 T
@ f (0 T ) = ; @T log B (0 T ) 0 T T :
Then f (t
(8.1)
and then the zero-coupon bond prices are determined by the initial conditions B (0 , gotten from the market, combined with the stochastic differential equation
T) 0
T
(8.3)
f Because all pricing of interest rate dependent assets will be done under the risk-neutral measure I , P f is a Brownian motion, we have written (8.1) and (8.3) in terms of W rather than f under which W W . Written this way, it is apparent that neither (t) nor (t T ) will enter subsequent computations. The only process which matters is (t T ) 0 t T T , and the process
(t T ) =
obtained from From (8.3) we see that (t T . Equation (8.4) implies This is because B (T vanish.
ZT
t
(t u) du
0 t T
(8.4)
(t T ).
T:
(8.5)
In conclusion, to implement the HJM model, it sufces to have the initial market data B (0 T T and the volatilities
T:
283
@T (t T )
ZT @ (t T ) = (t T ) ; (t t) = @u (t u) du:
t
f f We then let W be a Brownian motion under a probability measure I , and we let B (t T ) 0 t P T T , be given by (8.3), where r(t) is given by (8.2) and f (t T ) by (8.1). In (8.1) we use the initial conditions f f P Remark 28.3 It is customary in the literature to write W rather than W and IP rather than I , so that IP is the symbol used for the risk-neutral measure and no reference is ever made to the market measure. The only parameter which must be estimated from the market is the bond volatility (t T ), and volatility is unaffected by the change of measure.
@ f (0 T ) = ; @T log B (0 T ) 0 T T :
284
Chapter 29
Gaussian processes
Denition 29.1 (Gaussian Process) A Gaussian process X (t), t 0, is a stochastic process with the property that for every set of times 0 t1 t2 : : : tn , the set of random variables
is jointly normally distributed. Remark 29.1 If X is a Gaussian process, then its distribution is determined by its mean function
(s t) = IE (X (s) ; m(s)) (X (t) ; m(t))]: Indeed, the joint density of X (t 1) : : : X (tn) is IP fX (t1) 2 dx1 : : : X (tn) 2 dxng n o 1p exp ; 1 (x ; m(t)) 1 (x ; m(t))T dx1 : : : dxn = 2 (2 )n=2 det
;
where
::: 5 (tn t1) (tn t2 ) : : : (tn tn ) x1 x2 : : : xn], t is the row vector t1 t2 : : : tn ], and m(t) = m(t1) m(t2) : : : m(tn)]. IE exp
(t1 tn ) (t2 tn ) 7 7 7
(X n
where
u = u1 u2 : : : un].
k=1
uT
285
286
(s t) = s ^ t. Indeed, if 0 s t,
(s t) = IE W (s)W (t)] = IE W (s) (W (t) ; W (s)) + W 2 (s) = IEW (s):IE (W (t) ; W (s)) + IEW 2 (s) = IEW 2(s) = s ^ t:
To prove that a process is Gaussian, one must show that X (t 1) : : : X (tn) has either a density or a moment generating function of the appropriate form. We shall use the m.g.f., and shall cheat a bit by considering only two times, which we usually call s and t. We will want to show that
1 u1 u2 ] 2
"
11 21
12 22
# " #)
u1 u2
X (t) =
(s t) =
Proof: (Sketch.) We have Therefore,
Zs
0
2(u) du:
dX = dW:
deuX (s) = ueuX (s) (s) dW (s) + 1 u2euX (s) 2 (s) ds 2 Zs Zs uX (s) = euX (0) + u euX (v) (v ) dW (v ) + 1 u2 euX (v) 2 (v ) dv e 2 IEeuX (s) = 1 + 1 u2 2
| 0 Zs
0
Martingale
{z
d IEeuX (s) = 1 u2 2(s)IEeuX (s) 2 ds Zs uX (s) = euX (0) exp 1 u2 2(v ) dv IEe 2
= exp
2(v ) dv 1 u2 2 0
Zs
(1.1)
Rs This shows that X (s) is normal with mean 0 and variance 0 2(v ) dv .
287
s < t be given. Just as before, deuX (t) = ueuX (t) (t) dW (t) + 1 u2euX (t) 2 (t) dt: 2 Integrate from s to t to get
Now let 0
Zt
IE
Zt
s
Zs
0
(v )euX (v) dW (v )
to get
d uX (t) F (s) = 1 u2 2(t)IE euX (t) F (s) 2 dt IE e IE euX (t) F (s) = euX (s) exp 1 u2 2
Zt
s
2 (v )IE
Zt
s
2(v ) dv
X (t)), where 0 s t:
t 2 (1.2) (u1 +u2 )X (s) 1 (v ) dv = e exp 2 u2 2 s h i IE eu1 X (s)+u2 X (t) = IE IE eu1X (s)+u2 X (t) F (s)
t 2 1 u2 2 2 s (v ) dv Zs Zt (1.1) 2 (v ) dv + 1 u2 2 = exp 1 (u1 + u2 )2 2 2 2 s (v ) dv 0 Zs Zt 2 + 2u u ) 2(v ) dv + 1 u2 2 1 (u = exp 2 1 1 2 2 2 0 (v ) dv "R s 2 0R s 2# " #) ( 1 = exp 2 u1 u2 ] R0s 2 R0t 2 u1 : u2 0 0
IEX 2(s) =
Zs
0
2 (v ) dv
IE X (s)X (t)] =
Zs
0
IEX 2(t) =
Zt
0
2 (v ) dv
2(v ) dv:
288
Remark 29.2 The hard part of the above argument, and the reason we use moment generating functions, is to prove the normality. The computation of means and variances does not require the use of moment generating functions. Indeed,
X (t) =
is a martingale and X (0) = 0, so For xed s
Zt
0
(u) dW (u)
0,
Zs
0
2(v ) dv
s t,
Therefore,
IE 2(v ) dv and the same argument used above shows that for 0 s t, Zs 2(s) = IE X (s)X (t)] = IEX IE 2(v) dv: 0 However, when is stochastic, X is not necessarily a Gaussian process, so its distribution is not
0
determined from its mean and covariance functions. Remark 29.3 When is nonrandom,
IEX 2(s) =
Zs
X (t) =
Zt
0
(u) dW (u)
is also Markov. We proved this before, but note again that the Markov property follows immediately from (1.2). The equation (1.2) says that conditioned on F (s), the distribution of X (t) depends only R on X (s); in fact, X (t) is normal with mean X (s) and variance st 2(v ) dv .
289
z
z z
s y t (a)
y=
s v (b)
s v y t (c)
W (t) be a Brownian motion, and let (t) and h(t) be nonrandom functions. X (t) =
Zt
0
(u) dW (u)
Y (t) =
Zt
0
Then Y is a Gaussian process with mean function m Y (t) = 0 and covariance function
Y (s
t) =
Zs
0
2 (v )
Zs
v
h(y ) dy
Zt
v
h(y) dy dv:
(1.3)
Proof: (Partial) Computation of Y (s t): Let 0 s t be given. It is shown in a homework problem that (Y (s) Y (t)) is a jointly normal pair of random variables. Here we observe that
Zt
0
290 We have
Y (s
Z s0 Z t
0 Z sZ t 0
h(y)h(z )X (y)X (z ) dy dz
Zy
0 Z
2 (v ) dv dy dz
Zs
0
sZ s
0 y
h(y )h(z )
h(z )
Zt
z
Zy
0
2 (v ) dv
dy dz dz dy dz dy
(See Fig. 29.1(a))
Zz
0
2(v ) dv
Z sZ
0 Z 0
Zs
0 z
h(y )
Zs
y
h(z ) dz
Zy
0
2 (v ) dv
Z sZ
0 Z v
sZ y
h(z )
2 (v )
Zt
z
2(v ) dv
h(y ) dy dv dz
0 0 s h(z ) 2(v )
h(y )
2(v )
Zs
y
Zt
z
h(z ) dz dv dy h(z) dz dy dv
Zs
0
sZ s
h(y ) dy dz dv
0 v
Z0s 2 (v ) = 0 Zs 2 (v ) = 0
Zs
Zs
0
2 (v )
Z sZ t
v Z z
h(y ) 2(v )
sZ s
Zs
y
2 (v )
2 (v )
v y sZ t
Zv s v Zv
v s
h(y ) dy h(y ) dy
Zt Zv
v
h(z ) dz dv h(y ) dy dv
X (t) =
291
IE Y (t)jF (s)] =
Zs
0
s < t,
h(u)X (u) du + IE
Zt
s
= Y (s) + = Y (s) +
Zt Zt
s s
= Y (s) + X (s)
Zt
s
where we have used the fact that X is a martingale. The conditional expectation IE not equal to Y (s), nor is it a function of Y (s) alone.
292
Chapter 30
dr(t) = ( (t) ; (t)r(t)) dt + (t) dW (t) (t) are nonrandom functions of t. K (t) =
Zt
0
(u) du:
Then
eK (t)r(t) = r(0) +
so
Zt
0
eK(u) eK(u)
(u) du +
Zt
0
r(t) = e
K (t)
r(0) +
Zt
0
(u) du +
Zt
0
From Theorem 1.69 in Chapter 29, we see that r(t) is a Gaussian process with mean function
mr (t) = e
and covariance function
K (t)
r(0) +
;
Zt
0
^
eK(u) (u) du
(0.1)
r (s
t) = e
K (s) K (t)
Zs
0
(0.2)
293
294
RT We want to study 0
X (t) =
Then
Zt
0
ZT
0
ZT
0
r(t) = e r(t) dt =
ZT
0
r(0) +
K (t)
Zt
0
eK(u) (u) du + e
r(0) +
Zt
0
K (t) X (t)
eK (u) (u) du dt + Y (T ):
ZT
0
r(t) dt =
ZT
0
K (t)
Zt
0
eK (u) (u) du dt
(0.3)
var
ZT
0
ZT
0
e2K(v) 2(v )
ZT
v
K (y) dy
!2
dv:
( ZT ) B (0 T ) = IE exp ; r(t) dt 0 ( !) ZT ZT 1 (;1)2 var = exp (;1)IE r(t) dt + 2 r(t) dt 0 0 ZT ZTZ t = exp ;r(0) e K (t) dt ; e K (t)+K (u) (u) du dt 0 0 0 !2 ZT ZT
; ;
1 +2
= expf;r(0)C (0 T ) ; A(0 T )g
where
e2K (v) 2 (v )
K (y)
dy
dv
C (0 T ) = A(0 T ) =
ZT
0 0
Z TZt
0
K (t) dt
K (t)+K (u)
(u) du dt ; 1 2
ZT
0
ZT
v
K (y )
dy
!2
dv:
295
T u
Figure 30.1: Range of values of u
u
t for the integral.
30.1 Fiddling with the formulas
Note that (see Fig 30.1)
ZTZ t
0
= (y = t v = u) =
Therefore,
Z Z
0 TZ T 0 0
K (t)+K (u)
(u) du dt (u) dt du
K (y) dy
u T eK (v)
K (t)+K (u)
(v )
ZT
v
=
dv:
K (y)
ZT
v
!23 e K(y) dy 5 dv
;
"
Because r is a Markov process, this should be random only through a dependence on r(t). In fact,
296 where
C (t T ) = eK(t)
K (y) dy
; 1 e2K(v) 2(v) 2
ZT
v
!23 e K(y) dy 5 dv
;
K (y) dy:
The reason for these changes is the following. We are now taking the initial time to be t rather than RT R zero, so it is plausible that 0 : : : dv should be replaced by tT : : : dv: Recall that
K (v ) =
and this should be replaced by
Zv
0
(u) du
K (v ) ; K (t) =
Zv
t
(u) du:
Similarly, K (y ) should be replaced by K (y ) ; K (t). Making these replacements in A(0 see that the K (t) terms cancel. In C (0 T ), however, the K (t) term does not cancel.
T ), we
T ) and At (t T ) denote the partial derivatives with respect to t. From the formula
297
dB (t T ) = r(t)B (t T ) dt ; (t)C (t T )B(t T ) dW (t): In particular, the volatility of the bond price is (t)C (t T ).
30.3 Calibration of the Hull & White model
Recall:
K (y)
dy ; 1 e2K (v) 2 (v ) 2
ZT
v
!2 3 e K (y) dy 5 dv
;
T ) for all T 2 0 T ] from market data (with some interpolation). Can we (t), (t), and (t) for all t 2 0 T ]? Not quite. Here is what we can do.
;
B (0 T ) 0 T T r(0);
(0); (t) 0 t T
(0)C (0 T ) 0 T
C (0 T ) =
ZT
0
K (y)
dy:
We now have
B(0 T ) = expf;r(0)C (0 T ) ; A(0 T )g we can solve for A(0 T ) for all T 2 0 T ]. Recall that
K (y) dy
; 1 e2K(v) 2(v) 2
T
as follows:
ZT
v
ZT @ A(0 T ) = K (v) (v )e K (T ) ; e2K (v) 2 (v )e K (T ) e e K (y) dy dv @T 0 v !# Z " ZT K (T ) @ A(0 T ) = T eK (v) (v ) ; e2K (v) 2 (v ) K (y) dy e @T e dv 0 v @ eK (T ) @ A(0 T ) = eK (T ) (T ) ; Z T e2K(v) 2(v ) e K (T ) dv @T @T 0 @ eK (T ) @ A(0 T ) = e2K (T ) (T ) ; Z T e2K(v) 2 (v ) dv eK(T ) @T @T 0 @ eK(T ) @ eK (T ) @ A(0 T ) = (T )e2K (T ) + 2 (T ) (T )e2K(T ) ; e2K (T ) 2(T ) 0 T T : @T @T @T
; ; ; ; ; 0 0
(T ) 0 T
ZT"
!#
(t)e2K (t) + 2 (t) (t)e2K (t) ; e2K (t) 2(t) = known function of t:
From assumption 4 and step 1, we know all the coefcients in this equation. From assumption 3, we have the initial condition (0). We can solve the equation numerically to determine the function (t) 0 t T . Remark 30.1 The derivation of the ordinary differential equation for (t) requires three differentiations. Differentiation is an unstable procedure, i.e., functions which are close can have very different derivatives. Consider, for example,
299
g (x) = 10 cos(1000x)
0
jf (x) ; g (x)j = 10
0 0
Assumption 5 for the calibration was that we know the volatility at time zero of bonds of all maturities. These volatilities can be implied by the prices of options on bonds. We consider now how the model prices options.
R T1 r(u) du 0 (expf;r(T1)C (T1 T2) ; A(T1 T2)g ; K )+ : Z Z + = e x expf;yC (T1 T2) ; A(T1 T2)g ; K f (x y) dx dy RT where f (x y ) is the joint density of 0 1 r(u) du r(T1) . R We observed at the beginning of this Chapter (equation (0.3)) that 0T1 r(u) du is normal with "Z T1 # Z T1
= IEe
; 1 1 ; ;1 ;1
IE e
(B (T1 T2) ; K )
1 = IE
4
r(u) du =
=
Z T1
0
IEr(u) du r(0)e
;
2 = var 1
4
"Z T1
0
r(u) du =
# Z0T1
K (v) + e K (v)
;
Zv
0
Z T1
v
eK(u) (u) du dv
K (y) dy
!2
dv:
Z T1
0
300 In fact,
R T1 r(u) du r(T ) 1 0
1 2 = IE
"Z T1
0
= =
Z T1 Z T1
0 0
where r (u
;1
;1
1 p 2 1;
2 2 exp ; 2(1 ; 2) x2 + 2 xy + y 2 2 1 2 1 2
"
#)
dx dy:
(4.1)
0 T1] is
IE e
R T1 r(u) du
t
= IE e
R T1 r(u) du
t
(4.2)
Because of the Markov property, this is random only through a dependence on r(t). To compute R this option price, we need the joint distribution of tT1 r(u) du r(T1) conditioned on r(t). This
301
1 (t) = IE
"Z T1
t
Z T1
t
r(u) du F (t)
;
r(t)e
Zv
t
eK (u) (u) du dv
K (y) dy
dv
2 (t) = IE
r(T1) r(t)
2 (t) = IE 2
(r(T1) ;
2K (T1 )
=e
Z T1
t
(t))2
F (t) !
" Z T1
t
Z T1
The variances and covariances are not random. The means are random through a dependence on r(t). Advantages of the Hull & White model: 1. Leads to closed-form pricing formulas. 2. Allows calibration to t initial yield curve exactly. Short-comings of the Hull & White model: 1. One-factor, so only allows parallel shifts of the yield curve, i.e.,
B (t T ) = IE exp ;
ZT
t
r(u) du F (t)
can exceed 1.
302
Chapter 31
Cox-Ingersoll-Ross model
In the Hull & White model, r(t) is a Gaussian process. Since, for each t, r(t) is normally distributed, there is a positive probability that r(t) < 0. The Cox-Ingersoll-Ross model is the simplest one which avoids negative interest rates. We begin with a d-dimensional Brownian motion (W 1 W2 constants. For j = 1 : : : d, let Xj (0) 2 IR be given so that
: : : Wd ) .
Let
dXj (t) = ; 1 Xj (t) dt + 1 dWj (t): 2 2 Xj is called the Orstein-Uhlenbeck process. It always has a drift toward the origin. The solution to
this stochastic differential equation is
Xj (t) = e
1 2 t
Xj (0) + 1 2
;
Zt
0
e 2 u dWj (u) :
mj (t) = e
and covariance function
1 2 t Xj (0)
^
(s t) = 1 2 e 4
4
1 (s+t) Z s t u 2 e 0
du:
Dene
2 2 2 r(t) = X1 (t) + X2 (t) + : : : + Xd (t): 2 If d = 1, we have r(t) = X1 (t) and for each t, IP fr(t) > 0g = 1, but (see Fig. 31.1)
IP
There are innitely many values of t > 0 for which r(t) = 0 303
= 1
304
r(t) = X 1 (t)
2
t x2 ( X (t), X (t) ) 2 1
fxi xj = 2
0
d X i=1
if i = j if i 6= j:
It s formula implies o
dr(t) =
=
d X i=1 d X i=1
fxi dXi + 1 2
2Xi ; 1 Xi dt + 1 dWi(t) + 2 2
d X i !=1
dWi dWi
= ; r(t) dt +
2
Xi dWi + d4 dt
305
d X Xi dW = pr dWi
dW dW =
d X Xi2 i=1
i=1
r dt = dt
r(t) dW (t)
We dene
d = 4 2 > 0:
d X i=1
r(t) =
Xi2(t) <1 2
2), then
2 (i.e.,
1 2 ), then 2
d=
4 2 is a positive
0 be given. Take
306
(t t). Then
md(t) = e
d 1 X
;
1 q 2 t r(0)
r(t) =
(t t)
i=1
{z
X p i((tt)t)
4
!2 }
degrees of
+
term
;2
2 Xd (t) | {z }
(0.1)
As t!1, we have (t t) = 4 , and so the limiting distribution of r(t) is 4 times a chi-square with d = 4 2 degrees of freedom. The chi-square density with 4 2 degrees of freedom is
f (y ) =
We make the change of variable r = 4
1 2 = 2; 2
2
22
y=2 :
p(r) = 4 2 :
1 2 = 2; 2
2 2
4 r 2
= 22
22
;2 2
r:
307
-y
Figure 31.2: The function h(y )
Because we are going to apply the following analysis to the case X (t) 0 for all t.
0 at time 0. Then X (t) is random with density p(0 t x y ) (in the y We start at X (0) = x variable). Since 0 and x will not change during the following, we omit them and write p(t y ) rather than p(0 t x y ). We have
IEh(X (t)) =
for any function h.
h(y )p(t y ) dy
t and y . We derive it below. Let h(y ) be a smooth function of y 0 which vanishes near y = 0 and for all large values of y (see
Fig. 31.2). It s formula implies o
The Kolmogorov forward equation (KFE) is a partial differential equation in the forward variables
dh(X (t)) = h (X (t))b(X (t)) + 1 h (X (t)) 2(X (t)) dt + h (X (t)) (X (t)) dW (t) 2
0 00 0
so
Z th
0
0
Zt
0
Z th
0
308 or equivalently,
Z
0
Z tZ
1
Zt 1 2 0 0
Z0 0
h (y )b(y )p(s y) dy ds +
0
h (y ) 2 (y)p(s y) dy ds:
00
h(y)pt(t y) dy =
h (y )b(y )p(t y ) dy + 1 2
0
Z
0
h (y ) 2 (y)p(t y ) dy:
00
Z
0
y=
Z
0
h (y )
00
2(y )p(t
y) dy = h (y)
0
2 (y )p(t =0
=0
{z
y=0 } y=
;
1
Z
0
{z
y)
y=0 }
@ h (y ) @y
0
2(y )p(t
y ) dy
2(y )p(t
@ = ;h(y ) @y
2(y )p(t =0
{z
y)
y=
y=0 }
Z
0
@2 h(y) @y 2
y ) dy:
Therefore,
Z
0
h(y)pt(t y) dy = ;
Z
0
2(y )p(t
y ) dy
or equivalently,
Z
0
"
2 (y )p(t
y)
dy = 0:
This last equation holds for every function h of the form in Figure 31.2. It implies that
2 (y )p(t
y) = 0:
(KFE)
If there were a place where (KFE) did not hold, then we could take points, but take h to be zero elsewhere, and we would obtain
"
309
0 = t! pt (t y ): lim
1
@ (b(y)p(y )) ; 1 @ 2 2 @y 2 @y
2 (y )p(y )
= 0:
When an equilibrium density exists, it is the unique solution to this equation satisfying
p(y ) 0 8y 0
p(y ) dy = 1:
p(r) = Cr
where
;2 2
C = 22
We compute
22
p (r) = ; 22r2
00
We want to verify the equilibrium Kolmogorov forward equation for the CIR process:
(EKFE)
310 Now
@ (( ; r)p(r)) = ; p(r) + ( ; r)p (r) @r @ 2 ( 2 rp(r)) = @ ( 2 p(r) + 2rp (r)) @r2 @r = 2 2 p (r) + 2rp (r):
0 0 0 00
1 + r ( ; 1 2 ; r) + ; 2r ( ; 1 2 ; r)2 2 2 2 = p(r) ( ; 1 2 ; r) 2 ( ; 1 2 ; r) 2 2 2r ; 1 2 2 ( ; 1 2 ; r) =0 1 + r(
2 2 2r ; 1 2 ; r) ; 2 r ( 2 2 1 ; 2 2 ; r)2
as expected.
dr(t) = ( ; r(t)) dt +
where r(0) is given. The bond price process is
r(t) dW (t)
"
Because
"
the tower property implies that this is a martingale. The Markov property implies that B (t T ) is random only through a dependence on r(t). Thus, there is a function B (r t T ) of the three dummy variables r t T such that the process B (t T ) is the function B (r t T ) evaluated at r(t) t T , i.e.,
B(t T ) = B(r(t) t T ):
311
d exp ;
= exp ;
1 Brr (r(t) 2
The expression in
: : : ] equals
pr dW
0 t < T r 0:
(4.1)
B(r T T ) = 1 r 0: B (r t T ) = e
rC (t T ) A(t T )
;
Surprisingly, this equation has a closed form solution. Using the Hull & White model as a guide, we look for a solution of the form
;
where C (T
;1 ; Ct(t T ) + C (t T ) + 1 2C 2(t T ) = 0 C (T T ) = 0 2
and then set
A (t T ) =
ZT
t
C (u T ) du
312 so A(T
T ) = 0 and
At (t T ) = ; C (t T ):
C (t T ) =
sinh( (T ; t)) cosh( (T ; t)) + 1 sinh( (T ; t)) 2 2 3 1 2 log 4 e 2 (T t) 5 A(t T ) = ; 2 cosh( (T ; t)) + 1 sinh( (T ; t)) 2
;
where
1 2
2+2 2
u u sinh u = e ; e 2
;
u u cosh u = e + e : 2
;
"
where
B(r t T ) = exp f;rC (t T ) ; A(t T )g 0 t < T r 0 and C (t T ) and A(t T ) are given by the formulas above. Because the coefcients in dr(t) = ( ; r(t)) dt + r(t) dW (t) do not depend on t, the function B (r t T ) depends on t and T only through their difference = T ; t. Similarly, C (t T ) and A(t T ) are functions of = T ; t. We write B(r ) instead of B (r t T ), and we have B(r ) = exp f;rC ( ) ; A( )g 0 r 0
where
1 2
cosh( ) + 1 sinh( ) 2
3 5
2 + 2 2:
B (r(0) 0) = 1
313
Z
0
r(u) du = 0:
But also, so
B(r(0) T ) = exp f;r(0)C (T ) ; A(T )g r(0)C (0) + A(0) = 0 lim r(0)C (T ) + A(T )] = 1 T!
1
and
r(0)C (T ) + A(T )
is strictly inreasing in T .
"
where T1 is the expiration time of the option, T 2 is the maturity time of the bond, and 0 t T 1 n R o T2. As usual, exp ; 0t r(u) du v(t r(t)) is a martingale, and this leads to the partial differential equation (where v
Other European derivative securities on the bond are priced using the same partial differential equation with the terminal condition appropriate for the particular security.
dr(t) = ( ; r(t)) dt +
r(t) dW (t):
Real time scale: In this time scale, the interest rate r( t) is given by a time-dependent CIR equation ^^
t:
314
Process time
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
^ t = '(t)
^ t:
Real time
Figure 31.3: Time change function. There is a strictly increasing time change function t Fig. 31.3).
^rt ^ ^ ^ Let B (^ ^ T ) denote the price at real time t of a bond with maturity T when the interest rate at time ^ is r. We want to set things up so t ^ ^rt ^ B (^ ^ T ) = B (r t T ) = e
;
rC (t T ) A(t T )
;
^ where t = '(t)
^ T = '(T ), and C (t T ) and A(t T ) are as dened previously. ^ We need to determine the relationship between r and r. We have
^ ^ ^ ^ = '(T ), make the change of variable t = '(t), dt = ' (t) dt in the rst integral to get
0
315
31.7 Calibration
^ ^) ^ ^ ^r^ t ^ B (^(t) ^ T ) = B r(t^ '(t) '(T ) ' ) ( (t) ^ ^ C ('(t) '(T )) ; A('(t) '(T )) ^ ^ = exp ;r(t) ^^ ^ ' (t) n o = exp ;r(t)C (t T ) ; A(t T ) ^^^ ^ ^ ^^ ^
0 0
where
^ do not depend on ^ and T only through t are time dependent. ^r ^ Suppose we know r(0) and B (^(0)
Take and so the equilibrium distribution of r(t) seems reasonable. These values determine ^ the functions C A. Take '0(0) = 1 (we justify this in the next section). For each T , solve the ^): equation for '(T
^r ^ ^ ^ ^ ; log B (^(0) 0 T ) = r(0)C (0 '(T )) + A(0 '(T )): ^ The right-hand side of this equation is increasing in the '( T ) variable, starting at 0 at time having limit 1 at 1, i.e.,
(*)
0 and
^r ^ ^ ^ Since 0 ; log B (^(0) 0 T ) < 1 (*) has a unique solution for each T . For T ^1 < T2, then ^ is '(0) = 0. If T ^ ^ ^ ^ ; log B (r(0) 0 T1) < ; log B (r(0) 0 T2)
= 0, this solution
^ ^ so '(T1) < '(T2). Thus ' is a strictly increasing time-change-function with the right properties.
316
31.8 Tracking down '0 (0) in the time change of the CIR model
Result for general term structure models:
@ ; @T log B(0 T )
Justication:
T =0
= r(0):
T =0
In the real time scale associated with the calibration of CIR by time change, we write the bond price as
^r ^ B(^(0) 0 T )
thereby indicating explicitly the initial interest rate. The above says that
@T
T =0
The calibration of CIR by time change requires that we nd a strictly increasing function ' with '(0) = 0 such that
^r ^ ^ ^ where B (^(0) 0 T ), determined by market data, is strictly increasing in T , starts at 1 when T ^r ^ ^ and goes to zero as T !1. Therefore, ; log B (^(0) 0 T ) is as shown in Fig. 31.4.
Consider the function Here C (T ) and A(T ) are given by
(cal)
= 0,
r(0)C (T ) + A(T ) ^
C (T ) =
2 + 2 2:
317 Goes to 1
^r ^ 6; log B (^(0) 0 T )
Strictly increasing
-T ^
Figure 31.4: Bond price in CIR model
6
r(0)C (T ) + A(T ) ^
^r ^ ; log B (^(0) 0 T )
...... ...... ...... ...... ....... ....... ....... ....... ...... ...... ..... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
^ '(T )
Figure 31.5: Calibration
-T
^ The function r(0)C (T ) + A(T ) is zero at T = 0, is strictly increasing in T , and goes to 1 as T !1. This is because the interest rate is positive in the CIR model (see last paragraph of Section 31.4).
To solve (cal), let us rst consider the related equation
^r ^ ^ ^ ^ ; log B (^(0) 0 T ) = r(0)C ('(T )) + A('(T )): ^ ^ Fix T and dene '(T ) to be the unique T for which (see Fig. 31.5)
(cal)
dened a time-change function ' which has all the right properties, except it satises (cal) rather than (cal).
^r ^ ^ ; log B (^(0) 0 T ) = r(0)C (T ) + A(T ) ^ ^ ^ ^ ^ ^ ^ ^ If T = 0, then '(T ) = 0. If T1 < T2, then '(T1) < '(T2). As T !1, '(T )!1. We have thus
318 We conclude by showing that ' 0(0) = 1 so ' also satises (cal). From (cal) we compute
^r ^ r(0) = ; @^ log B (^(0) 0 T ) ^ ^ @T T =0 = r(0)C ('(0))' (0) + A ('(0))' (0) ^ = r(0)C (0)' (0) + A (0)' (0): ^
0 0 0 0 0 0 0 0
Note that r(0) is the initial interest rate, observed in the market, and is striclty positive. Dividing by ^ r(0), we obtain ^ ' (0) = 1:
0
Computation of C 0(0):
C( )=
0
1 cosh( ) + 1 sinh( ) 2
cosh( ) )+ 1 2
1 cosh( ) + 2 sinh( )
h i C (0) = 12 ( + 0) ; 0(0 + 1 ) = 1: 2
0
; sinh( )
2 sinh(
cosh( )
Computation of A0 (0):
A ( ) = ;2 2
0
"
cosh( ) + 1 sinh( ) 2
#
2
1 cosh( ) + 1 sinh( ) 2
=2
=2 2e
1 cosh( ) + 2 sinh( )
; e
A (0) = ; 2 2
0
=2
2 sinh(
+0
= ; 2 2 12 = 0:
1 1 ( + 0)2 #2 ( + 0) ; (0 + 2 ) " 2
1 2 ;2 2
1 )+ 2
cosh( )
Chapter 32
X1 (t) = Interest rate at time t X2(t) = Yield at time t on a bond maturing at time t +
Let X1(0) > 0, X2 (0) differential equations
> 0 be given, and let X1(t) and X2(t) be given by the coupled stochastic
dX1(t) = (a11X1(t) + a12X2(t) + b1) dt + 1 1X1(t) + 2X2(t) + dW1(t) (SDE1) q q dX2(t) = (a21X1(t) + a22X2(t) + b2) dt + 2 1X1(t) + 2X2(t) + ( dW1(t) + 1 ; 2 dW2(t))
(SDE2) where W1 and W2 are independent Brownian motions. To simplify notation, we dene
dW1(t) dW3(t) = dt
and
1 2Y
dt:
320
32.1 Non-negativity of Y
dY =
1 dX1 + 2 dX2 = ( 1a11X1 + 1a12 X2 + 1b1) dt + ( 2a21 X1 + 2a22 X2 + 2 b2) dt q p + Y ( 1 1 dW1 + 2 2 dW1 + 2 1 ; 2 2 dW2)
where
W4(t) = (
is a Brownian motion. We shall choose the parameters so that: Assumption 1: For some , Then
1 a11 + 2a21 =
1 a12 + 2a22 =
2:
dY =
1X1 +
+( = Y dt + (
dW4:
From our discussion of the CIR process, we recall that Y will stay strictly positive provided that: Assumption 2: and Assumption 3: Under Assumptions 1,2, and 3,
1b1 + 2 b2 ;
almost surely,
dX1(t) = (a11X1 (t) + a12X2 (t) + b1) dt + 1 Y (t) dW1(t) q dX2(t) = (a21X1 (t) + a22X2 (t) + b2) dt + 2 Y (t) dW3(t):
(SDE1) (SDE2)
321
"
Since the pair (X1 X2) of processes is Markov, this is random only through a dependence on X1(t) X2(t). Since the coefcients in (SDE1) and (SDE2) do not depend on time, the bond price depends on t and T only through their difference = T ; t. Thus, there is a function B (x 1 x2 ) of the dummy variables x1 x2 and , so that
"
exp ;
Zt
0
is a martingale. We compute its stochastic differential and set the dt term equal to zero.
d exp ;
= exp ; = exp ;
Zt
0
Z0t Zt
0
X1(u) du
+
+ 1 2 Y Bx1 x1 + 2 1
1
Y Bx1 dW1 +
dt
Y Bx2 dW3
x2 ) is ; x1B ; B +(a11x1 + a12x2 + b1)Bx1 +(a21x1 + a22x2 + b2)Bx2 + 1 2( 1x1 + 2x2 + )Bx1 x1 2 1 1 + 1 2( 1x1 + 2x2 + )Bx1 x2 + 2 2 ( 1x1 + 2x2 + )Bx2 x2 = 0: (PDE) 2 B (x1 x2 ) = exp f;x1 C1 ( ) ; x2C2( ) ; A( )g 0 and all x1 x2 satisfying 1x1 + 2 x2 + > 0:
(*)
B(x1 x2 0) = 1 8x1 x2 satisfying (*) because = 0 corresponds to t = T . This implies the initial conditions C1(0) = C2(0) = A(0) = 0: We want to nd C1( ) C2( ) A( ) for > 0. We have B (x1 x2 ) = ;x1 C1( ) ; x2C2( ) ; A ( ) B (x1 x2 ) Bx1 (x1 x2 ) = ;C1( )B(x1 x2 ) Bx2 (x1 x2 ) = ;C2( )B(x1 x2 ) 2 Bx1 x1 (x1 x2 ) = C1 ( )B(x1 x2 ) Bx1 x2 (x1 x2 ) = C1( )C2( )B (x1 x2 ) 2 Bx2 x2 (x1 x2 ) = C2 ( )B(x1 x2 ):
0 0 0
(IC)
(PDE) becomes
0 = B (x1 x2 ) ;x1 + x1C1( ) + x2 C2( ) + A ( ) ; (a11x1 + a12 x2 + b1)C1( ) ; (a21x1 + a22x2 + b2)C2( ) 2 + 1 2 ( 1x1 + 2 x2 + )C1 ( ) + 1 2( 1x1 + 2x2 + )C1( )C2( ) 2 1 2 + 1 2 ( 1x1 + 2 x2 + )C2 ( ) 2 2
0 0 0
+1 2 +1 2 +1 2
We get three equations:
0
2 1C 2 ( 1 1
)+
0
1 2 1C1(
)C2( ) + 1 2 )C2( ) + 1 2
2 2
2 1C 2 ( 2 2
) )
)+
1 2 2C1(
2 C 2( 2 2 2
C1 ( )C2( ) + 1 2
);
2 C2 ( )
2 C 2( 1 1 1
1 2 1C1 (
)C2( ) ; 1 2
2 C 2( 2 1 2
)
(1) (2) (3)
C1(0) = 0 2 C2( ) = a12C1( ) + a22 C2( ) ; 1 2 2C1 ( ) ; 2 1 C2(0) = 0 2 A ( ) = b1C1( ) + b2C2( ) ; 1 2 C1 ( ) ; 1 2 1 A(0) = 0
0
1 2 2C1( 2
)C2( ) ; 1 2
2 2
2 2C 2 ( 2 2 2 C2 ( )
C1 ( )C2( ) ; 1 2
323
We rst solve (1) and (2) simultaneously numerically, and then integrate (3) to obtain the function A( ).
32.3 Calibration
Let 0
0 is
This equation must hold for every value of X 1(t) and X2 (t), which implies that
C1( 0) = 0 C2 ( 0) =
We must choose the parameters
A( ) = 0:
1 2
1 2
324
Chapter 33
(0.1)
r(t) and the volatility process (t) are adapted to some ltration
(t) = exp
Zt
0
r(u) du
(t) F (t) (T )
326 so
0 = IE
(t) (S (T ) ; F (t T )) F (t) (T ) = (t)IE S (T ) Ft ; F (t T )IE (t) F (t) (T ) (T ) = (t) S (t) ; F (t T )B (t T ) (t) = S (t) ; F (t T )B (t T )
Therefore,
) F (t T ) = BSt(tT ) : (
Denition 33.1 (Num raire) Any asset in the model whose price is always strictly positive can be e taken as the num raire. We then denominate all other assets in units of this num raire. e e
Example 33.1 (Money market as num raire) The money market could be the num raire. At time t, the e e S (t) B (t T ) stock is worth (t) units of money market and the T -maturity bond is worth (t) units of money market.
Example 33.2 (Bond as num raire) The T -maturity bond could be the num raire. At time t e e is worth F(t T ) units of T -maturity bond and the T -maturity bond is worth 1 unit.
T , the stock
We will say that a probability measure IP N is risk-neutral for the num raire N if every asset price, e divided by N , is a martingale under IP N . The original probability measure IP is risk-neutral for the num raire (Example 33.1). e Theorem 0.71 Let N be a num raire, i.e., the price process for some asset whose price is always e strictly positive. Then IP N dened by
Z N (T ) dIP 8A 2 F (T ) (T )
A
327
IP and IPN are equivalent, i.e., have the same probability zero sets, and Z IP (A) = N (0) N((T )) dIPN 8A 2 F (T ): A T
1 Z N (T ) dIP IPN ( ) = N (0) (T ) 1 :IE N (T ) = N (0) (T ) (0) = N1 N(0) (0) =1
Proof: Because N is the price process for some asset, N= is a martingale under IP . Therefore,
Let Y be an asset price. Under IP , Y= is a martingale. We must show that under IPN , Y=N is a martingale. For this, we need to recall how to combine conditional expectations with change of measure (Lemma 1.54). If 0 t T T and X is F (T )-measurable, then
Therefore,
Y T Y IEN N(T ) F (t) = N((tt)) IE N((T )) N(T ) F (t) (T ) (T ) () = N((tt)) Y (tt) Y = N (t) (t) which is the martingale property for Y=N under IP N .
33.1 Bond price as num raire e
Fix T
2 (0 T ] and let B(t T ) be the num raire. The risk-neutral measure for this num raire is e e 1 Z B (T T )
IPT (A) = B (0 T ) dIP A (T ) Z 1 = 1 B (0 T ) A (T ) dIP 8A 2 F (T ):
328 Because this bond is not dened after time T , we change the measure only up to time T 1 using B (0 T ) B ((TT ) and only for A 2 F (T ). ) is called the stock is
T , i.e.,
IPT
T -forward measure.
Denominated in units of
) F (t T ) = BSt(tT ) (
0 t T:
dF (t T ) =
We write F (t) rather than F (t
F (t
T )F (t T ) dWT (t) 0 t T t
0.
(1.1)
i.e., a differential without a dt term. The process fWT 0 IPT . We may assume without loss of generality that F (t T )
Z S (T ) (T ) dIP 8A 2 F (T ):
A
(2.1)
F (t
T ) in (1.1).
In other
d F 1t) = (
(2.1)
329
d F 1t) = dg (F (t)) ( = g (F (t)) dF (t) + 1 g (F (t)) dF (t) dF (t) 2 1 = ; F 2 (t) F (t T )F (t T ) dWT (t) + F 31(t) 2 (t T )F 2(t T ) dt F
0 00
Under IPT ;WT is a Brownian motion. Under this measure, F 1t) has volatility F (t T ) and mean ( rate of return 2 (t T ). The change of measure from IPT to IPS makes F 1t) a martingale, i.e., it F ( changes the mean return to zero, but the change of measure does not affect the volatility. Therefore, (t T ) in (2.1) must be F (t T ) and WS must be
Zt
0
F (u
T ) du:
V (0) = IE (1T ) (S (T ) ; K )+ = IE S (T ) 1 S (T )>K ; KIE 1 1 S (T )>K (T ) Z(T ) S (T ) dIP ; KB (0 T ) Z 1 = S (0) S (0) (T ) B(0 T ) (T ) dIP S (T )>K S (T )>K = S (0)IPS fS (T ) > K g ; KB (0 T )IPT fS (T ) > K g = S (0)IPS fF (T ) > K g ; KB (0 T )IPT fF (T ) > K g = S (0)IPS 1 < 1 ; KB (0 T )IPT fF (T ) > K g: F (T ) K
f g f g f g f g
330 This is a completely general formula which permits computation as soon as we specify F (t we assume that F (t T ) is a constant F , we have the following:
T ). If
1 = B (0 T ) exp n W (T ) ; 1 2 T o F S 2 F F (T ) S (0) 1 S (0) IPS F (1T ) < K = IPS F WS (T ) ; 1 2 T < log KB (0 T ) 2 F p S (0) S p = IPS Wp(T ) < 1 log KB (0 T ) + 1 F T 2 = N ( 1) 1 p
where
1=
Similarly,
log
S (0) 1 2 KB (0 T ) + 2 F T :
o n F (T ) = BS (0) ) exp F WT (T ) ; 1 2 T F 2 (0 T IPT fF (T ) > K g = IPT F WT (T ) ; 1 2 T > log KB (0 T ) 2 F S (0) T 1 p = IPT Wp(T ) > 1 log KB (0 T ) + 2 2 T F S (0) T F T 1 p p = IPT ;WT (T ) < 1 log S (0) ; 2 2 T F KB(0 T ) T T F = N ( 2)
where
2=
If r is constant, then B (0
F
;
1 p
S (0) log KB (0 T ) ; 1 2 T : 2 F
T) = e
1=
rT ,
331
This formula also suggests a hedge: at each time t, hold KN ( 2(t)) bonds.
We want to verify that this hedge is self-nancing. Suppose we begin with $ V (0) and at each time t hold N ( 1(t)) shares of stock. We short bonds as necessary to nance this. Will the position in the bond always be ;KN ( 2(t))? If so, the value of the portfolio will always be S (t)N ( 1(t)) ; KB (t T )N ( 2(t)) = V (t) and we will have a hedge. Mathematically, this question takes the following form. Let
(t) = N ( 1(t)): At time t, hold (t) shares of stock. If X (t) is the value of the portfolio at time t, then X (t) ; (t (t)S (t) will be invested in the bond, so the number of bonds owned is X B)(t T )(t) S (t) and the
;
(3.1)
dV (t) = N ( 1(t)) dS (t) + S (t) dN ( 1(t)) + dS (t) dN ( 1(t)) ; KN ( 2(t)) dB(t T ) ; K dB(t T ) dN ( 2(t)) ; KB(t T ) dN ( 2(t)): If X (0) = V (0), will X (t) = V (t) for 0 t T ?
(3.2)
Formulas (3.1) and (3.2) are difcult to compare, so we simplify them by a change of num raire. e This change is justied by the following theorem. Theorem 3.73 Changes of num raire affect portfolio values in the way you would expect. e Proof: Suppose we have a model with k assets with prices S 1 S2 : : : Sk . At each time t, hold i (t) shares of asset i, i = 1 2 : : : k ; 1, and invest the remaining wealth in asset k. Begin with a nonrandom initial wealth X (0), and let X (t) be the value of the portfolio at time t. The number of shares of asset k held at time t is
dX =
=
Note that
k 1 X
;
k X i=1
i=1
i dSi + i dSi: k X
X;
k 1 X
;
i=1
i Si
dSk Sk
Xk (t) =
and we only get to specify Let N be a num raire, and dene e
:::
i (t)Si (t)
i = 1 2 : : : k:
1 1 b 1 dX = N dX + X d N + dX d N !
i i i
=
Now
i=1 k X i=1
Si i dc:
Therefore,
333
2 : : : k;
We return to the European call hedging problem (comparison of (3.1) and (3.2)), but we now use the zero-coupon bond as num raire. We still hold (t) = N ( 1(t)) shares of stock at each time t. e In terms of the new num raire, the asset values are e Stock: Bond: The portfolio value evolves according to
and
(3.2)
334
Chapter 34
Brace-Gatarek-Musiela model
34.1 Review of HJM under risk-neutral I P
f (t T ) = df (t T ) =
where The interest rate is r(t) = f (t
(t T ) (t T ) dt + (t T ) dW (t) (t T ) =
ZT
t
(t u) du
"
satisfy
dB (t T ) = r(t) B(t T ) dt ;
(t T )
t ) | ({z T }
B(t T ) dW (t):
0 t T:
where
> 0 is the constant volatility of the forward rate. This is not possible because it leads to
(t T ) =
( t T ) = f (t T )
ZT
t
f (t u) du
df (t T ) = 2f (t T )
ZT
t
f (t u) du dt + f (t T ) dW (t)
335
336 and Heath, Jarrow and Morton show that solutions to this equation explode before T .
The problem with the above equation is that the dt term grows like the square of the forward rate. To see what problem this causes, consider the similar deterministic ordinary differential equation
f (t) = f 2(t)
0
= T ; t:
(2.1) (2.2)
D(t ) = B(t t + )
= exp ; (u = v ; t du = dv ) : = exp ; = exp ;
Z t+ Zt Z0
0
(2.3)
f (t v ) dv
f (t t + u) du r(t u) du
(2.4)
337
(2.5) (2.6)
@ @
(t ) =
Z
0
(t u) du
(2.7) (2.8)
dr(t ) =
(2.5),(2.2) (2.8)
df (t {z+ ) | t }
@ + @T f (t t + ) dt
(t ) (t ) dt + (t ) dW (t) + @@ r(t ) dt
(2.9)
Also,
dD(t ) =
(2.6),(2.4)
= r(t) B (t t + ) dt ; (t )B (t t + ) dW (t) ; r(t )D(t ) dt (2.1) = r(t 0) ; r(t )] D(t ) dt ; (t )D(t ) dW (t): (2.10)
34.3 LIBOR
Fix > 0 (say, time t + . L(t
1 = 4 year). $ D(t ) invested at time t in a (t + )-maturity bond grows to $ 1 at 0) is dened to be the corresponding rate of simple interest:
338
) is dened to be the simple (forward) interest rate for this investment: D(t + ) (1 + L(t )) = 1 D(t ) R D(t ) = exp f; 0 r(t u) dugo n R 1 + L(t ) = D(t + ) exp ; 0 + r(t u) du = exp
(Z nR
r(t u) du
L(t ) =
Connection with forward rates:
exp
r(t u) du ; 1
(4.1)
( @ exp Z @
so
r(t u) du
)
=0
(Z
r(t u) du
)
=0
nR + #0
exp
nR
r(t u) du ; 1 >0
xed: (4.2)
r(t u) du ; 1
r(t ) is the continuously compounded rate. L(t ) is the simple rate over a period of duration . We cannot have a log-normal model for r(t ) because solutions explode as we saw in Section 34.1. For xed positive , we can have a log-normal model for L(t ).
34.5 The dynamics of L(t )
We want to choose
339
0 0. This is the BGM model, and is a subclass of HJM models, for some (t ) t corresponding to particular choices of (t ).
Recall (2.9):
r(t u) du =
=
! Z Z
+ +
dr(t u) du
(5.1)
and
r(t u) du d
r(t u) du
+
r(t u) du
!2
(5.2)
340 But
2 nR @ L(t ) = @ 4 exp @ @ (Z +
= exp
r(t u) du ; 1 5
Then
(5.4)
) (t + ) = (t ) + L(t L(t(t ) ) : 1+
Plugging this into (5.4) yields
(5.3)
"
(5.4)
L(0 )
(t )
0 0:
from market data. Choose a forward LIBOR volatility function (usually nonrandom)
t 0
341
Because LIBOR gives no rate information on time periods smaller than , we must also choose a partial bond volatility function
(t )
With these functions, we can for each
t 0 0
<
L(t ) t 0 0
Plugging the solution into (5.3), we obtain
(t ) for
L(t ) t 0
and we continue recursively.
Remark 34.1 BGM is a special case of HJM with HJMs (t In BGM, (t ) is usually taken to be nonrandom; the resulting
Remark 34.2 (5.4) (equivalently, (5.4)) is a stochastic partial differential equation because of the @ @ L(t ) term. This is not as terrible as it rst appears. Returning to the HJM variables t and T , set
K (t T ) = L(t T ; t):
Then
(t )L(t ) = 1 + L(t )] (t + ) ; (t ) :
If we let
#0, then
@ (t )L(t )! @
(t + )
=0
= (t )
L(t )!r(t ) = f (t t + )
342 so
K (t T )!f (t T ):
Remark 34.4 Although the dt term in (6.1) has the term to this equation do not explode because
2(t
T ; t)K 2 (t T ) 1 + K (t T )
(t) = exp
nR t
WT (t) = W (t) +
is a Brownian motion under IPT .
Zt
0
(u T ; u) du
0 t T
343
K (t T ) = K (0 T ) exp
and
Zt
0
(u T ; u) dWT + (u) ; 1 2
Zt
0
2(u
T ; u) du
K (T T ) = K (0 T ) exp
= K (t T ) exp
We assume that
(Z T
0
(Z T
t
(u T ; u) dWT + (u) ;
ZT 2(u 1 2 0 ZT
t
2 (u
T ; u) du
)
(8.1)
(u T ; u) dWT + (u) ; 1 2
T ; u) du :
is nonrandom. Then
X (t) =
is normal with variance
ZT
t
(u T ; u) dWT + (u) ; 1 2
2(t) =
ZT
t
2(u
T ; u) du
(8.2)
ZT
t
2(u
T ; u) du
Case I: T
t T+
(9.1)
t T.
IPT + (A) =
where
Z
A
Z (T + ) dIP
8A 2 F (T + )
T Z (t) = (B (t (0 + +) ) : t)B T
We have
Z (T + )
K (T T ) = K (t T ) expfX (t)g R 1 where X (t) is normal under IPT + with variance 2(t) = tT 2(u T ; u) du and mean ; 2 2(t). Furthermore, X (t) is independent of F (t). CT + (t) = B(t T + )IET + (K (t T ) expfX (t)g ; c)+ F (t) :
Set
1 log y ; 1 (t) : ( t) c 2
(9.2)
CT + (t) = B(t T + ) g (K (t T )) 0 t T ; :
In the case of constant , we have
(t) =
T ;t
345
T0 = 0 T1 = T2 = 2 : : : Tn = n :
(K (Tk Tk ) ; c)+
at time Tk+1
k = 0 1 : : : n ; 1:
The value at time t of the cap is the value of all remaining caplets, i.e.,
C (t) =
34.11 Calibration of BGM
The interest rate caplet c on L(0
k:t Tk
CTk (t):
T ) at time T +
CT + (0) = B(0 T + ) g (K (0 T ))
where g (dened in the last section) depends on
ZT
0
2 (u
T ; u) du:
Let us suppose
(t ) = ( ):
Then g depends on
ZT
0
2(T ; u) du =
ZT
0
2(v ) dv:
Z T0
0
2(v ) dv
Z T1
T0
2(v ) dv =
Z T0
0
2(v ) dv
:::
In this case, we may assume that is constant on each of the intervals
;
Z Tn
Tn;1
Tn )
346
R (0) for all T 0, we can back out 0T 2(v ) dv and then differenp 2( ) for all 0. )= 0, and To implement BGM, we need both ( )
(t )
and choose these constants to make the above integrals have the values implied by the caplet prices.
t 0 0
< : t+
(see (2.6)).
Now (t ) is the volatility at time t of a zero coupon bond maturing at time < , it is reasonable to set Since is small (say 1 year), and 0 4
(t ) = 0
We can now solve (or simulate) to get
t 0 0
0
< :
L(t ) t 0
or equivalently,
K (t T ) t 0 T 0
D(t n ) = exp
= =
n Y k=1 n Y k=1
(Z n
0
exp
(Z k
r(t u) du
(k 1)
;
r(t u) du
1 + L(t (k ; 1) )]
where the last equality follows from (4.1). The long rate is
T1 = T0 +
T2 = T0 + 2 : : : Tn = T0 + n :
347
(L(Tk 0) ; c)
at time Tk+1
k = 0 1 : : : n ; 1:
Now
1 + L(Tk 0) = B (T 1T ) k k+1
so
We compute
B (Tk Tk+1 )
n 1 X
;
= B (t T0) ; (1 + c)B (t T1) + B (t T1) ; (1 + c)B (t T2) + : : : + B (t Tn 1 ) ; (1 + c)B (t Tn ) = B (t T0) ; cB (t T1) ; cB (t T2) ; : : : ; cB (t Tn ) ; B (t Tn ): The forward swap rate wT0 (t) at time t for maturity T0 is the value of c which makes the time-t
;
k=0
B (t Tk ) ; (1 + c)B(t Tk+1)]
In contrast to the cap formula, which depends on the term structure model and requires estimation of , the swap formula is generic.
348
Chapter 35
350
351
The rst derivation of the Black-Scholes formula given in this course, using only It s formula, o is similar to that originally given by Black & Scholes (1973). An important companion paper is Merton (1973), which makes good reading even today. (This and many other papers by Merton are collected in Merton (1990).) Even though geometric Brownian motion is a less than perfect model for stock prices, the Black-Scholes option hedging formula seems not to be very sensitive to deciencies in the model.
35.7 Girsanovs theorem, the martingale representation theorem, and risk-neutral measures.
Girsanovs Theorem in the generality stated here is due to Girsanov (1960), although the result for constant was established much earlier by Cameron & Martin (1944). The theorem requires a f technical condition to ensure that IEZ (T ) = 1, so that I is a probability measure; see Karatzas & P Shreve (1991), page 198. The form of the martingale representation theorem presented here is from Kunita & Watanabe (1967). It can also be found in Karatzas & Shreve (1991), page 182. The application of the Girsanov Theorem and the martingale representation theorem to risk-neutral pricing is due to Harrison & Pliska (1981). This methodology frees the Brownian-motion driven model from the assumption of constant interest rate and volatility; these parameters can be random through dependence on the path of the underlying asset, or even through dependence on the paths of other assets. When both the interest rate and volatility of an asset are allowed to be stochastic, the Brownian-motion driven model is mathematically the most general possible for asset prices without jumps.
352 When asset processes have jumps, risk-free hedging is generally not possible. Some works on hedging and/or optimization in models which allow for jumps are Aase (1993), Back (1991), Bates (1988,1992), Beinert & Trautman (1991), Elliott & Kopp (1990), Jarrow & Madan (1991b,c), Jones (1984), Madan & Seneta (1990), Madan & Milne (1991), Mercurio & Runggaldier (1993), Merton (1976), Naik & Lee (1990), Schweizer (1992a,b), Shirakawa (1990,1991) and Xue (1992). The Fundamental Theorem of Asset Pricing, as stated here, can be found in Harrison & Pliska (1981, 1983). It is tempting to believe the converse of Part I, i.e., that the absence of arbitrage implies the existence of a risk-neutral measure. This is true in discrete-time models, but in continuous-time models, a slightly stronger condition is needed to guarantee existence of a risk-neutral measure. For the continuous-time case, results have been obtained by many authors, including Stricker (1990), Delbaen (1992), Lakner (1993), Delbaen & Schachermayer (1994a,b), and Fritelli & Lakner (1994, 1995). In addition to the fundamental papers of Harrison & Kreps (1979), and Harrison & Pliska (1981, 1983), some other works on the relationship between market completeness and uniqueness of the risk-neutral measure are Artzner & Heath (1990), Delbaen (1992), Jacka (1992), Jarrow & Madan (1991a), M ller (1989) and Taqqu & Willinger (1987). u
353
and by Parkinson (1977), Johnson (1983), Geske & Johnson (1984), MacMillan (1986), Omberg (1987), Barone-Adesi & Whalley (1987), Barone-Adesi & Elliott (1991), Bunch & Johnson (1992), Broadie & Detemple (1994), and Carr & Faguet (1994).
354
35.14 REFERENCES
(1993) Aase, K. K., A jump/diffusion consumption-based capital asset pricing model and the equity premium puzzle, Math. Finance 3(2), 6584. (1961) Alexander, S. S., Price movements in speculative markets: trends or random walks, Industrial Management Review 2, 726. Reprinted in Cootner (1964), 199218. (1990) Artzner, P. & Heath, D., Completeness and non-unique pricing, preprint, School of Operations Research and Industrial Engineering, Cornell University. (1900) Bachelier, L., Th orie de la sp culation, Ann. Sci. Ecole Norm. Sup. 17, 2186. Reprinted e e in Cootner (1964). (1991) Back, K., Asset pricing for general processes, J. Math. Econonics 20, 371395. (1991) Barone-Adesi, G. & Elliott, R., Approximations for the values of American options, Stoch. Anal. Appl. 9, 115131. (1987) Barone-Adesi, G. & Whalley, R., Efcient analytic approximation of American option values, J. Finance 42, 301320. (1988) Bates, D. S., Pricing options under jump-diffusion processes, preprint, Wharton School of Business, University of Pennsylvania. (1992) Bates, D. S., Jumps and stochastic volatility: exchange rate processes implicit in foreign currency options, preprint, Wharton School of Business, University of Pennsylvania. (1991) Beinert, M. & Trautmann, S., Jump-diffusion models of stock returns: a statistical investigation, Statistische Hefte 32, 269280. (1984) Bensoussan, A., On the theory of option pricing, Acta Appl. Math. 2, 139158. (1983) Biger, H. & Hull, J., The valuation of currency options, Financial Management, Spring, 2428. (1986) Billingsley, P., Probability and Measure, 2nd edition, Wiley, New York. (1975) Black, F., Fact and fantasy in the use of options, Financial Analysts J., 31, 3641, 6172.
355
(1990) Black, F., Derman, E. & Toy, W., A one-factor model of interest rates and its application to treasury bond options, Fin. Anal. Journal, 3339. (1973) Black, F. & Scholes, M., The pricing of options and corporate liabilities, J. Political Economy 81, 637659. (1964) Boness, Some evidence of the protability of trading put and call options. In Cootner (1964), 475496. (1994) Bouaziz, L., Bryis, E. & Crouhy, M., The pricing of forward-starting Asian options, J. Banking Finance. (1994a) Brace, A. & Musiela, M., Swap derivatives in a Gaussian HJM framework, preprint, School of Mathematics, University of New South Wales. (1994b) Brace, A. & Musiela, M., A multi-factor Gauss Markov implementation of Heath, Jarrow and Morton, Math. Finance 4, 259283. (1995) Brace, A.. Gatarek, D. & Musiela, M., The market model of interest rate dynamics, preprint, School of Mathematics, University of New South Wales. (1977) Brennan, M. J. & Schwartz, E. S., The valuation of the American put option, J. Finance 32, 449462. (1979) Brennan, M. J. & Schwartz, E. S., A continuous-time approach to the pricing of bonds, J. Banking Finance 3, 133155. (1982) Brennan, M. J. & Schwartz, E. S., An equilibrium model of bond pricing and a test of market efciency, J. Financial Quantitative Analysis 17, 301329. (1994) Broadie, M. & Detemple, J., American option values: new bounds, approximations and a comparison of existing methods, preprint, Graduate School of Business, Columbia University. (1992) Bunch, D. & Johnson, H., A simple and numerically efcient valuation method for American puts, using a modied Geske-Johnson approach, J. Finance 47, 809816. (1944) Cameron, R. H. & Martin, W. T., Transformation of Wiener integrals under translations,
356 Ann. Math. 45, 386396. (1994) Carr, P. & Faguet, D., Fast accurate valuation of American options, preprint, Johnson School of Business, Cornell University. (1995) Carverhill, A. & Pang, K., Efcient and exible bond option valuation in the Heath, Jarrow, and Morton framework, J. Fixed Income, September, 7077. (1968) Chung, K. L. A Course in Probability Theory, Academic Press, Orlando, FL. (1983) Chung, K. L. & Williams, R. J. (1983) Introduction to Stochastic Integration, Birkh user, a Boston. (1991) Conz , A. & Viswanathan, R., Path dependent options: the case of lookback options, J. e Finance 46, 18931907. (1964) Cootner, P. H. (editor), The Random Character of Stock Market Prices, MIT Press, Cambridge, MA. (1981) Cox, J. C., Ingersoll, J. E., & Ross, S, The relation between forward prices and futures prices, J. Financial Economics 9, 321346. (1985a) Cox, J. C., Ingersoll, J. E. & Ross, S., An intertemporal general equilibrium model of asset prices, Econometrica 53, 363384. (1985b) Cox, J. C., Ingersoll, J. E. & Ross, S., A theory of the term structure of interest rates, Econometrica 53, 373384. (1976) Cox, J. C. & Ross, S. A., The valuation of options for alternative stochastic processes, J. Fin. Economics 3, 145166. (1979) Cox, J. C., Ross, S. A. & Rubinstein, M., Option pricing: a simplied approach, J. Financial Economics 7, 229263. (1985) Cox, J. C. & Rubinstein, M. Options Markets, Prentice Hall, Englewood Cliffs, NJ. (1986) DeGroot, M., Probability and Statistics, 2nd edition, Addison Wesley, Reading, MA. (1992) Delbaen, F., Representing martingale measures when asset prices are continuous and bounded, Math. Finance 2(2), 107130.
357
(1994a) Delbaen, F. & Schachermayer, W., A general version of the fundamental theorem of asset pricing, Mathematische Annalen 300, 463520. (1994b) Delbaen, F. & Schachermayer, W., Arbitrage and free-lunch with bounded risk, for unbounded continuous processes, Math. Finance 4(4), 343348. (1953) Doob, J., Stochastic Processes, Wiley, New York. (1990) Dothan, M. U., Prices in Financial Markets, Oxford University Press, New York. (1989) Dufe, D., Futures Markets, Prentice-Hall, Englewood Cliffs, NJ. (1992) Dufe, D., Dynamic Asset Pricing Theory, Princeton University Press, Princeton. (1993) Dufe, D. & Kan, R., A yield-factor model of interest rates, preprint. (1994) Dufe, D. & Kan, R., Multi-factor term structure models, Phil. Transactions Royal Society London A 347, 577586. (1992) Dufe, D. & Stanton, R., Pricing continuously resettled contingent claims, J. Economic Dynamics and Control, 16, 561573. (1905) Einstein, A. On the movement of small particles suspended in a stationary liquid demanded by the molecular-kinetic theory of heat, Ann. Physik 17. (1990) Elliott, R. J. & Kopp, P. E., Option pricing and hedge portfolios for Poisson processes, Stoch. Anal. Appl. 9, 429444. (1965) Fama, E. The behavior of stock-market prices, J. Business 38, 34104. (1948) Feynman, R., Space-time approach to nonrelativistic quantum mechanics, Rev. Mod. Phys. 20, 367387. (1966) Fisk, D. L., Sample quadratic variation of continuous, second-order martingales, Z. Wahrscheinlichkeitstheorie verw. Gebiete 6, 273278. (1995) Flesaker, B. & Hughston, L., Positive interest, preprint. (1993) F llmer, H. & Schweizer, M., A microeconomic approach to diffusion models for stock o prices, Math. Finance 3, 123.
358 (1994) Fritelli, M. & Lakner, P., Almost sure characterization of martingales, Stochastics 49, 181 190. (1995) Fritelli, M. & Lakner, P., Arbitrage and free-lunch in a general nance market model: the fundamental theorem of asset pricing, IMA Vol 65: Mathematical Finance, 8992, (M. Davis, D. Dufe, W. Fleming & S. Shreve, eds.), Springer-Verlag, New York. (1983) Garman, M. & Kohlhagen, S., Foreign currency option values, J. International Money Finance 2, 231237. (1995) Geman, H., El Karoui, N. & Rochet, J.-C., Changes of num raire, changes of probability e measure, and option pricing, J. Applied Probability 32, 443458. (1993) Geman, H. & Yor, M., Bessel processes, Asian options, and perpetuities, Mathematical Finance 3, 349375. (1984) Geske, R. & Johnson, H., The American put-option valued analytically, J. Finance 39, 1511 1524. (1960) Girsanov, I. V., On transforming a certain class of stochastic processes by absolutely continuous substitution of measures, Theory Probability Applications 5, 285301. (1979) Goldman, B., Sosin, H. & Gatto, M., Path dependent options: Buy at the low, sell at the high, J. Finance 34, 6984. (1979) Goldman, B., Sosin, H. & Shepp, L. A., On contingent claims that insure ex-post optimal stock market timing, J. Finance 34, 401414. (1979) Harrison, J. M. & Kreps, D. M., Martingales and arbitrage in multiperiod securities markets, J. Economic Theory 20, 381408. (1981) Harrison, J. M. & Pliska, S. R., Martingales and stochastic integrals in the theory of continuous trading, Stochastic Processes Applications 11, 215260. (1983) Harrison, J. M. & Pliska, S. R., A stochastic calculus model of continuous trading: complete markets, Stochastic Processes Applications 15, 313316. (1990) Heath, D., Jarrow, R. & Morton, A., Bond pricing and the term structure of interest rates: a discrete time approximation, J. Financial Quantitative Analysis 25, 419440. (1992) Heath, D., Jarrow, R. & Morton, A., Bond pricing and the term structure of interest rates: a
359
(1993) Hogan, M., Problems in certain two-factor term structure models, Annals Applied Probability 3, 576581. (1986) Ho, T. S. Y. & Lee, S.-B., Term structure movements and pricing interest rate contingent claims, J. Finance 41, 10111029. (1988) Huang, C.-f. & Litzenberger, R., Foundations for Financial Economics, North Holland, Amsterdam. (1993) Hull, J., Options, Futures, and other Derivative Securities, Prentice Hall, Englewood Cliffs, NJ. (1990) Hull, J. & White, A., Pricing interest rate derivative securities, Rev. Financial Studies 3, 573592. (1994a) Hull, J. & White, A., Numerical procedures for implementing term structure models I: single-factor models, J. Derivatives, Fall, 716. (1994b) Hull, J. & White, A., Numerical procedures for implementing term structure models II: two factor models, J. Derviatives, Winter. (1987) Ingersoll, J. E., Theory of Financial Decision Making, Rowman & Littleeld. (1944) It , K., Stochastic integral, Proc. Imperial Acad. Tokyo 20, 519524. o (1992) Jacka, S., A martingale representation result and an application to incomplete nancial markets, Math. Finance 2, 239250. (1990) Jaillet, P., Lamberton, D. & Lapeyre, B., Variational inequaltiies and the pricing of American options, Acta Appl. Math. 21, 263289. (1989) Jamshidian, F., An exact bond option pricing formula, J. Finance 44, 205209. (1990) Jamshidian, F., The preference-free determination of bond and option prices from the spot interest rate, Advances Futures & Options Research 4, 5167. (1993a) Jamshidian, F., Price differentials, Risk Magazine 6/7.
360 (1993b) Jamshidian, F., Corralling quantos, Risk Magazine 7/3. (1996) Jamshidian, F., Sorting out swaptions, Risk Magazine 9/3, 5960. (1988) Jarrow, R., Finance Theory, Prentice Hall, Englewood Cliffs, NJ. (1991a) Jarrow, R. & Madan, D., A characterization of complete security markets on a Brownian ltration, Math. Finance 1(3), 3143. (1991b) Jarrow, R. & Madan, D., Valuing and hedging contingent claims on semimartingales, preprint, Johnson School of Business, Cornell University. (1991c) Jarrow, R. & Madan, D., Option pricing using the term structure of interest rates to hedge systematic discontinuities in asset returns, preprint, Johnson School of Business, Cornell University. (1981) Jarrow, R. & Oldeld, G., Forward contracts and futures contracts, J. Financial Economics 9, 373382. (1983) Johnson, H., An analytic approximation for the American put price, J. Financial Quant. Analysis 18, 141148. (1984) Jones, E. P., Option arbitrage and strategy with large price changes, J. Financial Economics 13, 91113. (1951) Kac, M., On some connections between probability theory and differential and integral equations, Proc. 2nd Berkeley Symposium on Math. Stat. & Probability, 189215. (1988) Karatzas, I., On the pricing of American options, App. Math. Optimization 17, 3760. (1991) Karatzas, I. & Shreve, S. E., Brownian Motion and Stochastic Calculus, Springer-Verlag, New York. (1981) Karlin, S. & Taylor, H. M., A Second Course in Stochastic Processes, Academic Press, Orlando, FL. (1953) Kendall, M. G., Analysis of economic time-series Part I: prices, J. Royal Statistical Society 96, Part I, 1125. Reprinted in Cootner (1964), 8599. (1992) Kloeden, P. E. & Platen, E., Numerical Solution of Stochastic Differential Equations, SpringerVerlag, New York.
361
(1933) Kolmogorov, A. N., Grundbegriffe der Wahrscheinlichkeitsrechnung, Ergebnisse Mathematik 2. English translation: Foundations of Probability Theory, Chelsea, New York, 1950. (1967) Kunita, H. & Watanabe, S., On square-integrable martingales, Nagoya Math. Journal 30, 209245. (1992) Kunitomo, N. & Ikeda, M., Pricing options with curved boundaries, Mathematical Finance 2, 275298. (1993) Lakner, P., Martingale measures for a class of right-continuous processes, Math. Finance 3, 4353. (1993) Lamberton, D., Convergence of the critical price in the approximation of American options, Math. Finance 3, 179190. (1989) LeRoy, S. F., Efcient capital markets and martingales, J. Economic Literature 27, 1583 1621. (1939) L vy, P., Sur certains processus stochastiques homog nes, Compositio Math. 7, 283339. e e (1948) L vy, P., Processus Stochastiques et Mouvement Brownian, Gauthier-Villars, Paris. e (1992a) Longstaff, F. A. & Schwartz, E. S., Interest rate volatility and the term structure: a twofactor general equilibrium model, J. Finance 47, 12591282. (1992b) Longstaff, F. A. & Schwartz, E. S., A two-factor interest rate model and contingent claim valuation, J. Fixed Income 3, 1623. (1986) MacMillan, L., Analytic approximation for the American put option, Adv. Futures Options Research 1, 119139. (1990) Madan, D. & Seneta, E., The V. G. model for share returns, J. Business 63, 511524. (1991) Madan, D. & Milne, F., Option pricing with V. G. martingale components, Math. Finance 1(4), 3956. (1976) Margrabe, W., A theory of forward and futures prices, preprint, Wharton School of Business, University of Pennsylvania. (1965) McKean, H. P., Jr., Appendix to P. A. Samuelson (1965): A free boundary problem for the heat equation arising from a problem in mathematical economics, Industrial Management Review
362 6, 3239. (1993) Mercurio, F. & Runggaldier, W., Option pricing for jump diffusions: approximations and their interpretation, Math. Finance 3, 191200. (1969) Merton, R. C., Lifetime portfolio selection under uncertainty: the continuous-time case, Review Economics Statistics 51, 247257. (1973) Merton, R. C., Theory of rational option pricing, Bell J. Econ. Management Science 4, 141 183. (1976) Merton, R. C., Option pricing when underlying stock returns are discontinuous, J. Fin. Economics 3, 125144. (1990) Merton, R. C., Continuous-Time Finance, Basil Blackwell, Cambridge, 1990. (1989) M ller, S. On complete securities markets and the martingale property on securities prices, u Economic Letters 31 3741. (1996) Musiela, M. & Rutkowski, M., Arbitrage Pricing of Derivative Securities. Theory and Applications, Springer-Verlag, New York, to appear. (1992) Myneni, R., The pricing of the American option, Ann. Applied Probability 2, 123. (1990) Naik, V. & Lee, M., General equilibrium pricing of options on the market portfolio with discontinuous returns, Rev. Financial Studies 3, 493521. (1995) ksendal, B., Stochastic Differential Equations, 4th Edition, Springer-Verlag, New York. (1987) Omberg, E., The valuation of American puts with exponential exercise policies, Adv. Futures Option Research 2, 117142. (1959) Osborne, M. F. M., Brownian motion in the stock market, Operations Research 7, 145173. Reprinted in Cootner (1964), 100128. (1977) Parkinson, M., Option pricing: the American put, J. Business 50, 2136. (1990) Protter, P., Stochastic Integration and Differential Equations, Springer-Verlag, New York. (1995) Reed, N., If the cap ts : : : , Risk 8(8), 3435.
363
(1981) Richard, S. F. & Sundaresan, M., A continuous-time equilibrium model of forward and futures prices in a multigood economy, J. Financial Economics, 9, 347371. (1987) Ritchken, P. Options: Theory, Strategy and Applications, Harper Collins. (1995) Rogers, L. C. G., Which model of term-structure of interest rates should one use?, IMA Vol. 65: Mathematical Finance, 93116, (M. Davis, D. Dufe, W. Fleming & S. Shreve, editors), Springer-Verlag, New York. (1995) Rogers, L. C. G. & Shi, Z., The value of an Asian option, J. Applied Probability, to appear. (1991) Rubinstein, M. & Reiner, E., Breaking down the barriers, Risk Magazine 4/8, 2835. (1965) Samuelson, P. A., Proof that properly anticipated prices uctuate randomly, Industrial Management Review 4, 4150. (1973) Samuelson, P. A., Mathematics of speculative prices, SIAM Review 15, 142. (1992a) Schweizer, M., Martingale densities for general asset prices, J. Math. Economics 21, 123 131. (1992b) Schweizer, M., Mean-variance hedging for general claims, Annals Appl. Probab. 2, 171 179. (1990) Shirakawa, H., Optimal dividend and portfolio decisions with Poisson and diffusion-type return processes, preprint IHSS 90-20, Tokyo Institute of Technology. (1991) Shirakawa, H., Interest rate option pricing with Poisson-Gaussian forward rate curve processes, Math. Finance 1(4), 7794. (1961) Sprenkle, C. M., Warrant prices as indicators of expectations and preferences, Yale Economic Essays 1, 178231. Reprinted in Cootner (1964), 412474. (1990) Stricker, C., Arbitrage et lois de martingales, Ann. Inst. henri Poincar 26, 451460. e (1987) Taqqu, M. & Willinger, W., The analysis of nite security markets using martingales, Adv. Appl. Probab. 19, 125. (1977) Vasicek, O., An equilibrium characterization of the term structure, J. Financial Economics 5, 177188.
364 (1994) Vetzal, K. R., A survey of stochastic continuous time models of the term structure of interest rates, Insurance: Mathematics and Economics 14, 139161. (1939) Ville, J., Etude Critique de la Notion du Collectif, Gauthier-Villars, Paris. (1923) Wiener, N., Differential spaces, J. Math. Physics 2, 131174. (1924) Wiener, N., Un probl me de probabilit s d nombrables, Bull. Soc. Math. France 52, 569 e e e 578. (1991) Williams, D., Probability with Martingales, Cambridge University Press, Cambridge. (1993) Wilmott, P., Dewynne, J. & Howison, S., Option Pricing, Oxford Financial Press, PO Box 348, Oxford, OX4 1DR, UK. (1995) Wilmott, P., Dewynne, J. & Howison, S., The Mathematics of Financial Derivatives, Cambridge University Press, Cambridge. (1992) Xue, X. X., Martingale representation for a class of processes with independent increments, and its applications, in Lecture Notes in Control and Information Sciences 177, 279311, SpringerVerlag, Berlin.